阅读说明：本技术 基于人脸识别的车辆安全报警系统 (Vehicle safety alarm system based on face recognition ) 是由 王斐 孟洪兵 朱彩蝶 田慧娟 夏热扎提·阿地力 于 2021-09-15 设计创作，主要内容包括：本发明公开了一种基于人脸识别的车辆安全报警系统,包括雷达检测单元,摄像头采集及图像处理模块,蓝牙模块,手机或车载电子通讯设备,预警设备和云服务器,通过雷达检测单元作为预警监测,在满足雷达预警条件时启动拍照程序进行拍照,对采集的驾驶员状态图像数据,不仅要对图像处理,而且需要参与大数据学习,这些图像数据特征与困倦特征匹配程度高,图像数据量较少且集中,处理量较小；该系统在大部分时间处于摄像头无启动运行阶段,降低能耗和节约空间资源,提高设备使用寿命,在wifi环境利用云服务器平台进行特定驾驶员包括面部特征等深度学习,以及通过后台技术人员对图像数据特征进行标注,以强化学习。(The invention discloses a vehicle safety alarm system based on face recognition, which comprises a radar detection unit, a camera acquisition and image processing module, a Bluetooth module, a mobile phone or vehicle-mounted electronic communication equipment, early warning equipment and a cloud server, wherein the radar detection unit is used for early warning monitoring, a photographing program is started to photograph when a radar early warning condition is met, and acquired driver state image data need to be processed and participate in big data learning; the system is in a camera non-starting operation stage in most of time, reduces energy consumption, saves space resources, prolongs service life of equipment, utilizes the cloud server platform to perform deep learning of specific drivers including facial features and the like in a wifi environment, and labels image data features through background technicians to reinforce learning.)

1. A vehicle safety alarm system based on face recognition is characterized by comprising a radar detection unit, a camera acquisition and image processing module, a Bluetooth module, a mobile phone or vehicle-mounted electronic communication device, an early warning device and a cloud server, wherein the radar detection unit comprises a support, a power supply module, a radar module and a wireless transmission module, the radar device is fixed at a proper position in a vehicle through the support, a radar probe part is compared with a cab area, the camera acquisition unit comprises a support, a camera, a Bluetooth module and a processor, the radar detection unit is used for early warning monitoring, a photographing program is started to photograph when radar early warning conditions are met, acquired driver state image data is not only required to be processed, but also required to participate in big data learning, and the acquired image data is data acquired under necessary conditions, therefore, the image data features are matched with the drowsiness features to a high degree, the image data amount is small and concentrated, and the processing amount is small; the photographed image and the mobile phone Bluetooth are transmitted, the mobile phone sends image information to a background cloud server for big data analysis, the analysis result is transmitted to a mobile phone APP in real time, the mobile phone sends an alarm signal, a camera acquisition unit is used for monitoring the image in detail, only special images are sent wirelessly, the acquired image is sent only characteristic images during driving, the data transmission flow is reduced, the feedback speed is improved, a large amount of data are transmitted under a WIFI environment, a relation is established with a cloud server platform, deep learning and labeled data training are carried out through the cloud server platform, a training model is fed back to the mobile phone APP or vehicle-mounted equipment, the mobile phone or the vehicle-mounted equipment utilizes a pre-training model as a comparison basis for processing of the image data of a driver under the WIFI environment, deep learning including facial features and the like of a specific driver is carried out under the WIFI environment by utilizing the cloud server platform, and image data features are labeled by background technicians, to enhance learning.

2. The vehicle safety alarm system based on face recognition of claim 1, wherein the radar detection unit and the camera acquisition unit establish a signal transmission relationship in a wired or wireless manner, and the radar detection unit only needs to provide a starting time for the camera acquisition unit, so that the radar detection unit and the camera acquisition unit can be respectively arranged at different positions in the vehicle.

3. The vehicle safety alarm system based on face recognition of claim 1, wherein the camera acquisition unit sends image information to an APP terminal of the mobile phone or the vehicle-mounted mobile electronic device through Bluetooth to store corresponding image data, the APP terminal of the mobile phone or the vehicle-mounted mobile electronic device is connected with the cloud server platform to upload big data and download a training model, and the mobile phone or the vehicle-mounted mobile electronic device transmits corresponding model data to the camera acquisition unit to facilitate detection and judgment of a vehicle running state.

4. The vehicle safety alarm system based on face recognition of claim 1, wherein the warning device warns the driver through mobile phone warning or vibration or electric pulse through a cushion or a bracelet.

5. The vehicle safety warning system based on face recognition according to claim 1, wherein when the driver is in a static state t (1) and when the vehicle is driven for a cumulative t (2), the camera captures detailed state information of the driver's face and processes the image to determine whether the driver is in a tired or sleeping state.

6. The vehicle safety alarm system based on the face recognition is characterized in that the autonomous radar device comprises a base (1), a universal adjusting seat (2), a filter cover (3) and a radar device (4), and a display device, the radar device (4) is used for acquiring the action change information of the driver, and triggering the camera acquisition unit in parallel when capturing the action information, but no action information exists in the time t (1), the radar equipment triggers the camera acquisition unit to start to work cooperatively, the direction of the probe of the radar equipment is firstly adjusted to enable the probe to face a driver, secondly, the area outside the driver is shielded by the filtering wave cover (3), only the signal capture is carried out within the range of the driver, and finally, according to the body position of the driver, the driver is adaptively adjusted and tracked through the universal adjusting seat (2), so that the radar equipment is only used for automatically tracking and detecting the area near the driver.

7. The vehicle safety alarm system based on the face recognition is characterized in that the universal adjusting base (2) comprises a first plate layer (5), a second plate layer (6) and a third plate layer (7), a left-right spring (8) is fixed at the right end between the first plate layer (5) and the second plate layer (6), and a front-back spring (9) is fixed at the rear end between the second plate layer (6) and the third plate layer (7), so that the first plate layer (5) and the second plate layer (6) can rotate left and right relatively, the second plate layer (6) and the third plate layer (7) can rotate front and back relatively, and the combination of the left-right rotation and the front-back rotation enables the first plate layer (5) and the third plate layer (7) to rotate front and back, left and right to a proper position within a certain range; a left-right adjusting motor (10) is fixed at the outer end of the first plate layer (5), an adjusting screw is connected to a rotating shaft of the motor, and the adjusting screw is connected into a corresponding screw hole on the second plate layer (6); a front and back adjusting motor (11) is fixed at the outer end of the first plate layer (5) in a excluding way, an adjusting screw is connected on a rotating shaft of the motor, the adjusting screw penetrates through a through hole (42) arranged on the second plate layer (6) and then is connected in a corresponding screw hole on the third plate layer (7) to respectively control the left and right adjusting motor (10) and the front and back adjusting motor (11) to rotate, so that the relative left and right rotation of the first plate layer (5) and the second plate layer (6) can be realized, the relative front and back rotation of the second plate layer (6) and the third plate layer (7) can be realized, a controller automatically controls the rotation degree of the two motors according to a feedback signal of radar equipment, further automatically controls the first plate layer (5) and the third plate layer (7) to rotate to a proper position in a certain range in a front and back left and right direction, and the radar equipment (4) is installed at the center of the third plate layer (7), the filtering cover (3) is sleeved on the outer side of the radar device (4).

8. The vehicle safety alarm system based on the face recognition is characterized in that the filter cover (3) is arranged to be of an axially telescopic adjusting structure, a support (15) is fixed at the outer end of the third plate layer (7), the support (15) comprises a fixed sleeve, a telescopic filter cover (16) is sleeved in the fixed sleeve, an axial limiting block (17) is arranged on the inner side of the fixed sleeve, the outer side of the telescopic filter cover (16) is provided with threads, the threads are also provided with axial clamping grooves (18) along the axial direction, the axial limiting blocks (17) are embedded into the axial clamping grooves (18), the outer side of the fixed sleeve is sleeved with a rotatable adjusting sleeve (19), the adjusting sleeve (19) and the outer side of the telescopic filtering cover (16) are sleeved together, the adjusting sleeve (19) is in threaded connection with the telescopic filtering cover (16), and the adjusting sleeve (19) can be rotated to drive the telescopic filtering cover (16) to stretch axially.

9. The vehicle safety alarm system based on the face recognition is characterized in that a through hole is formed in the center of a third plate layer (7) and is provided with the radar device (4), a movable clamping seat (21) is arranged on a rear end mounting seat (20) of a radar transmitting and receiving probe, a static clamping seat (24) is fixed at the through hole, a sawtooth-shaped convex-concave engagement structure is arranged on the butt joint surface between the movable clamping seat (21) and the static clamping seat (24), a sleeve (22) connected with the center of the movable clamping seat (21) penetrates through the through hole and is provided with a disc spring (23), so that the movable clamping seat (21) and the static clamping seat (24) have an elastic compression relationship, and the rotary radar device can change the angular position; the power supply and the signal wire of the radar equipment are led out from the sleeve (22) and then connected with the circuit board (25), and the circuit board (25) is fixed between the board layers or in the base.

10. The vehicle safety alarm system based on the face recognition is characterized in that the connecting seat (27) is connected with the base (28) through the rotating shaft (35), so that the connecting seat (27) can rotate relative to the base (28), adjusting arms (29) are hinged between the connecting seat (27) and the base (28) through pin shafts respectively, the tail ends of the two adjusting arms (29) are fixed through lock pins (30), and the connecting seat (27) has an inclination angle relative to the base (28).

Technical Field

The invention belongs to the technical field of vehicle-mounted face recognition equipment, and particularly relates to a vehicle safety alarm system based on face recognition.

Background

At present, with the continuous development of economy, various automobiles become the most important transportation means for people to go out daily, and on the other hand, with the increase of the vehicle keeping amount in the global range, the rapid increase of the number of traffic accidents has become a serious social problem. According to relevant data statistics, about one third of casualties in traffic accidents every year are caused by fatigue driving. The data of the national statistical bureau show that the number of the traffic accidents per year in nearly five years in China exceeds 12 thousands, wherein the truck traffic accidents are particularly serious, the rate of the truck accidents is higher than that of the common motor vehicles, and the caused loss is also higher than the average level. Among them, traffic accidents due to fatigue driving cause significant losses to people's life and property safety every year, and various studies have shown that about 20% of all road accidents are associated with fatigue, and up to 50% on some roads.

Because of drowsiness and drowsiness caused by long-time driving or insufficient sleep, driving fatigue is one of the important causes of malignant traffic accidents as shown by data in many countries; the driver fatigue is the same as that of drunk driving and becomes a main hidden danger of traffic accidents, but the driver fatigue is easy to detect after drunk driving and has certain concealment. Driving fatigue refers to a brief and involuntary loss of attention, often characterized by a loss of consciousness and frequent and involuntary closure of both eyes. The existing fatigue driving detection is mainly used for detecting human face eye characteristics and abnormal behaviors of a driver and making judgment by combining a machine learning method, and because the fatigue driving detection of the method needs to collect a large amount of data and different driving habits of the driver have larger deviation, the real-time performance and the accuracy of the program are still required to be improved.

With the high development of intelligent sensors, intelligent pattern recognition, automotive electronics and vehicle dynamics technologies, the fatigue driving detection device and technology become a research hotspot in the field of traffic safety at home and abroad in recent years. Through the detection of the fatigue state of the driver, traffic accidents caused by fatigue driving can be greatly reduced. Particularly, for drivers who are engaged in business operations such as long-distance passenger transport, freight transport and the like, due to professional requirements, the drivers often drive continuously for a long time, and the drivers are difficult to keep a high-alert state during driving; it is therefore more important to detect the fatigue state in real time. The existing fatigue driving detection device, such as a fatigue driving early warning method and device with the publication number of 109902663A, comprises the steps of S1, acquiring depth image data and color image data of a driver in real time; step S2, recognizing key face information of the driver based on the color image data; step S3, recognizing body posture information of the driver based on the depth image data; and step S4, determining whether the driver has fatigue driving according to the key information of the human face and the body posture information. The technology is a fatigue driving detection device and method based on visual characteristics, the technical means is that a computer visual technology is used for extracting visual characteristic information of a driver to carry out fatigue judgment, the visual characteristics are greatly influenced by illumination, turning and line changing can greatly influence detection based on vehicle behaviors, and the accuracy, reliability and practicability are low.

A method, a system, a storage medium and an automobile for early warning simulation of abnormal driving state in a vehicle-displaying mode, wherein the method is published with number 111422206 a, and the method comprises the following steps: when the vehicle is in the vehicle exhibition mode, the vehicle-mounted system is in a simulated driving state in response to the input simulated driving instruction; in the simulated driving state, shooting a simulated image of a user at the main driving position in real time, acquiring simulated driving monitoring data, and sending the simulated driving monitoring data to a vehicle-mounted system interface for displaying; monitoring the abnormal driving state of a user at a main driving position through the simulation image, and marking the abnormal simulation image corresponding to the abnormal driving state; and displaying the marked abnormal simulation image and the corresponding simulation driving monitoring data on a vehicle-mounted system interface. The technical means of the scheme is to utilize a vehicle-mounted sensor to detect characteristics of the vehicle such as speed, lateral acceleration, transverse displacement, lane departure, change of a vehicle running track and the like so as to estimate the fatigue state of a driver. However, the accuracy of such a fatigue driving detection method is yet to be further improved due to limitations imposed by the specific model of the vehicle, the specific conditions of the road, and the individual driving habits, driving experiences, and driving conditions of the driver.

In a fatigue driving reminding system with publication number CN 106408874 a, an information station collects action information of a driver to be reminded in a driving process and a heart rate curve of the driver to be reminded, and sends the action information and the heart rate curve to a communication terminal. And the communication terminal receives the action information and the heart rate curve and sends the action information and the heart rate curve to the server. The server receives the action information and the heart rate curve, judges whether the driver to be reminded is tired or not according to the action information, the heart rate curve, the average energy and the accident heart rate curve, and if yes, the server sends an alarm signal to the communication terminal. The technology is a fatigue driving detection device and method based on the physiological characteristics of a driver, and the technical means is to use some physiological index sensors to judge whether the driver enters a fatigue state. Because related researches show that the physiological response of a human body becomes dull in a fatigue state, namely, the response of the physiological signal of the human body is delayed, and the index also deviates from a normal state.

The prior face recognition detection technology or the driver physiological index detection technology generally has the technical problems that: 1) the physiological index signal of a human body is weak, and the fatigue driving detection precision is not high due to the interference of movement and environmental noise on signal acquisition in the driving process of a driver; 2) most of the existing fatigue driving detection devices based on physiological characteristics adopt intelligent bracelets, utilize various sensors (such as skin temperature sensors, heart rate sensors, photoplethysmography sensors and the like) and various biological signals to detect the fatigue driving state, but still can not monitor the brain and eye movement functions, and the sensitivity and specificity are still not ideal enough; 3) most of the existing fatigue driving detection methods based on physiological characteristics adopt a preset comparison threshold value mode for judgment (if the heart rate is smaller than a certain set threshold value, fatigue driving is judged), and the difference of the individual physical conditions of drivers causes that the threshold value comparison method has no universality; 4) the existing equipment applied to fatigue driving generally takes camera acquisition as a main part, so that the problems of overlarge image processing information quantity, high energy consumption and storage space requirements and the like are caused, and the problems of higher processing performance and high cost of the equipment required to be configured are solved. Because the fatigue state is a limit state and an accidental event, the real-time camera shooting and image processing are adopted, so that the equipment is in an invalid starting operation stage for most of time, resources are wasted, and the service life of the equipment is influenced.

Disclosure of Invention

The invention provides a vehicle safety alarm system based on face recognition, which aims at solving the problems that the prior fatigue driving equipment generally takes camera acquisition as a main part, so that the image processing information amount is too large, the energy consumption and the storage space requirement are high, and the like, and the prior fatigue driving early warning equipment has high sensitivity and is easy to bring special interference to a driver.

The technical scheme adopted by the invention for solving the technical problems is as follows: the utility model provides a vehicle safety alarm system based on face identification, this system includes radar detection unit, camera collection and image processing module, bluetooth module, cell-phone or on-vehicle electronic communication equipment, early warning equipment and cloud ware, radar detection unit includes the support, power module, radar module (G fine motion does radar response module) and wireless sending module, radar equipment passes through the support to be fixed in the car in suitable position, and with the regional driver's cabin of radar probe part contrast, can be a plurality of to a plurality of radar module. The camera acquisition unit comprises a support, a camera, a Bluetooth module and a processor, a photographing program is started to photograph when the radar early warning condition is met by taking the radar detection unit as early warning monitoring, acquired driver state image data are not only processed, but also need to participate in big data learning, and the acquired image data are acquired under necessary conditions, so that the matching degree of the image data characteristics and the drowsiness characteristics is high, the image data amount is small and centralized, and the processing amount is small; the photographed image and the mobile phone Bluetooth are transmitted, the mobile phone sends image information to a background cloud server for big data analysis, the analysis result is transmitted to a mobile phone APP in real time, the mobile phone sends an alarm signal, a camera acquisition unit is used for monitoring the image in detail, only special images are sent wirelessly, the acquired image is sent only characteristic images during driving, the data transmission flow is reduced, the feedback speed is improved, a large amount of data are transmitted under a WIFI environment, a relation is established with a cloud server platform, deep learning and labeled data training are carried out through the cloud server platform, a training model is fed back to the mobile phone APP or vehicle-mounted equipment, the mobile phone or the vehicle-mounted equipment utilizes a pre-training model as a comparison basis for processing of the image data of a driver under the WIFI environment, deep learning including facial features and the like of a specific driver is carried out under the WIFI environment by utilizing the cloud server platform, and image data features are labeled by background technicians, to enhance learning. A background provides a limited typical model of experience E, reinforcement learning random acquisition is carried out, or a large amount of image data which are acquired at a special moment and are not marked are uploaded to a cloud platform in a network environment, the cloud platform marks a newly uploaded image automatically according to a large data sample of an image of other people marked artificially, and the newly uploaded image is downloaded to a vehicle-mounted machine end to be used as a model sample for field direct comparison.

The radar detection unit and the camera acquisition unit establish a signal transmission relation in a wired or wireless mode, and the radar detection unit only needs to provide starting time for the camera acquisition unit, so that the radar detection unit and the camera acquisition unit can be respectively arranged at different positions in the vehicle.

The camera acquisition unit send image information to cell-phone or on-vehicle mobile electronic device's APP end through the bluetooth, make corresponding image data obtain the storage, establish with cloud server platform through cell-phone or on-vehicle mobile electronic device's APP end and be connected, realize big data upload and the download of training model to and transmit corresponding model data to camera acquisition unit so that carry out detection and judgement under the vehicle driving state through cell-phone or on-vehicle mobile electronic device.

The warning device warns the driver through a mobile phone or by vibrating or electrically pulsing through a cushion or a bracelet.

The camera captures detailed status information of the driver's face while the driver is in a static state t and accumulates t while driving the vehicle, and processes the image to determine whether the driver is in a tired or sleeping state.

The automatic control radar equipment comprises a base, a universal adjusting seat, a filter cover and radar equipment, and display equipment, the radar equipment is used for acquiring action change information of a driver, the camera acquisition unit is triggered when the action information is captured, but no action information exists in t time, the radar equipment triggers the camera acquisition unit to start to work in a matching way, the direction of a probe of the radar equipment is firstly adjusted, the radar equipment faces the driver, the area outside the driver is shielded by the filter cover, signals are captured only in the range of the driver, and finally the driver is tracked by self-adaptive adjustment of the universal adjusting seat according to the body position of the driver, so that the radar equipment is only used for automatically tracking and detecting the area near the driver.

The universal adjusting seat comprises a first plate layer, a second plate layer and a third plate layer, a left-right direction elastic sheet is fixed at the right end part between the first plate layer and the second plate layer, and a front-back direction elastic sheet is fixed at the rear end part between the second plate layer and the third plate layer, so that the first plate layer and the second plate layer can rotate left and right relatively, the second plate layer and the third plate layer can rotate front and back relatively, and the combination of left and right rotation and front and back rotation enables the first plate layer and the third plate layer to rotate front and back and left and right to proper positions within a certain range; the outer end of the first plate layer is fixed with a left-right adjusting motor, a rotating shaft of the motor is connected with an adjusting screw, and the adjusting screw is connected in a corresponding screw hole in the second plate layer. The outer end of the first plate layer is also fixedly provided with a front and back adjusting motor, an adjusting screw is connected to a rotating shaft of the motor, the adjusting screw penetrates through a through hole formed in the second plate layer and is connected into a corresponding screw hole formed in the third plate layer, the left and right adjusting motor is controlled to rotate and the front and back adjusting motor is controlled to rotate respectively, the first plate layer and the second plate layer can rotate left and right relatively, the second plate layer and the third plate layer rotate front and back relatively, a controller automatically controls the rotation degree of the two motors according to a radar equipment feedback signal, and then the first plate layer and the third plate layer can rotate to a proper position in a certain range front and back left and right directions under automatic control, the radar equipment is installed at the center of the third plate layer, and the filtering cover is sleeved outside the radar equipment.

But set up the filter mantle and adjust the structure for the axial is flexible, be fixed with the support at the third plate layer external end, this support includes a fixed sleeve, and this fixed sleeve endotheca is equipped with flexible filter mantle, is provided with the axial stopper in the fixed sleeve inboard with getting rid of, sets up the screw thread in the flexible filter mantle outside, has the axial draw-in groove along the axial depression on the screw thread with getting rid of, the embedding of axial stopper in the axial draw-in groove fixed sleeve outside cover is equipped with can the pivoted adjust the cover, adjust the cover with get rid of the suit with threaded connection between flexible filter mantle, adjust the cover and the flexible filter mantle, rotate and adjust the cover and can drive flexible filter mantle and stretch out and draw back along the axial.

The center of the third plate layer is provided with a through hole and the radar equipment is installed, the rear end installation seat of the radar transmitting and receiving probe is provided with a movable clamping seat, the position of the through hole is fixedly provided with a static clamping seat, a sawtooth-shaped convex-concave connection structure is arranged on a butt joint surface between the movable clamping seat and the static clamping seat, a sleeve connected with the center of the movable clamping seat penetrates through the through hole and is provided with a disc spring, so that the movable clamping seat and the static clamping seat have an elastic compression relation, and the radar equipment can be rotated to change the angular position. The power supply and the signal wire of the radar equipment are led out from the sleeve and then connected with the circuit board, and the circuit board is fixed between the board layers or in the base.

Be connected through the pivot between connecting seat and base to the connecting seat can rotate for the base, has the regulating arm through round pin axle hinged respectively between connecting seat and base with getting rid of, and the end of two regulating arms is fixed through the lockpin, makes the connecting seat have an inclination for the base.

The invention has the beneficial effects that:

the system solves the problems that the prior fatigue driving equipment generally takes camera acquisition as a main part, so that the image processing information quantity is overlarge, the energy consumption and storage space requirements are high, and the like, and can reduce the equipment configuration requirement to reduce the cost. The system is in the camera non-starting operation stage in most time, so that the energy consumption is reduced, the space resource is saved, and the service life of the equipment is prolonged.

Because radar detection unit only needs to provide the start opportunity with camera acquisition unit, so preferably establish the communication through the bluetooth for radar detection unit and camera acquisition unit can be arranged in different positions in the car respectively. Unlike camera acquisition units that must be used to acquire facial features of the driver, radar detection units may be placed, for example, on the right side, front side, rear side, top, left side, or in hidden areas of the driver. Therefore, the arrangement modes of the radar detection units are more flexible and diversified.

For the collected driver state image data, not only the image processing but also the large data learning are required, and since the collected image data is data acquired under the necessary conditions, the degree of matching of the image data features with the drowsiness features is high, and the amount of image data is small and concentrated, so the processing amount is small. The system is characterized by comprising a large number of transmission devices under a WIFI environment, only characteristic images are sent during travel, data transmission flow is reduced, feedback speed is increased, the system needs to establish a relationship with a cloud server platform, deep learning and training of labeled data are carried out through the cloud server platform, a training model is fed back to the vehicle-mounted device, the vehicle-mounted device utilizes an early-stage training model as a driver image data processing comparison basis under a WIFI-free environment, deep learning of specific drivers including facial features and the like is carried out by utilizing the cloud server platform under the WIFI environment, and labeling of the image data features is carried out through background technicians, so that reinforcement learning is achieved.

The apparatus performs image capturing and image processing work only under necessary conditions. Most of the time, the driver is in a non-fatigue state, and only the radar equipment with low energy consumption and low running cost is used for monitoring operation. The conditions for triggering the camera to start work and image processing can be controlled by setting sensitivity, but the radar is mainly used for detecting the human motion state, namely the radar acquires the head and neck motion state, the arm motion state, the body motion state and the like of a driver (other passengers) so as to determine that the driver is in a waking state. The above is the selection characteristic of the intervention time of the camera acquisition unit, so that the relevance of the image information of the camera acquisition unit after intervention or driving and the drowsiness characteristic is strong, a large amount of irrelevant image information is eliminated, and the information amount and the cost of image processing are greatly simplified.

For the collected driver state image data with high association degree, big data learning needs to be participated in further, and the feature data which is preliminarily screened by the radar module and has high association degree with drowsiness can simplify the marking difficulty of big data features and improve the data training precision.

The system avoids the defects of low information processing speed, delayed feedback result and the like caused by real-time background communication, and can remarkably reduce data transmission flow and improve feedback speed by acquiring the training model and carrying out image processing under the condition of not giving background direct communication, thereby realizing the effect of light artificial intelligent image processing. The timing of establishing the relation between the system and the cloud server platform can be in any other non-driving state, deep learning and labeled data training are carried out through the cloud server platform, a training model is fed back to the vehicle-mounted equipment, and image data features are labeled through background technicians, so that learning is strengthened.

The method comprises the steps of adopting the automatic control radar equipment, firstly adjusting the direction of a probe of the radar equipment to enable the probe to face a driver, secondly shielding the area outside the driver through a filter cover, and only capturing signals within the range of the driver, and finally adaptively adjusting and tracking the driver through a universal adjusting seat according to the body position of the driver, so that the radar equipment is only used for automatically tracking and detecting the area near the driver. The radar equipment can independently operate, and the controller controls the left and right adjusting motors to rotate and the front and back adjusting motors to rotate according to real-time feedback signals of the radar equipment, so that the orientation of the radar equipment can be adaptively changed and positioned along with the body position of a driver, and the action information of the driver can be obtained.

When being applied to universal regulation seat with the universal screw axle, can provide first sheet layer, second sheet layer and third sheet layer and have more rotation function to can improve rotation range for the embodiment, make this radar check out test set can be placed in and be close driver's position more.

The warning device is a bracelet, electromagnets are respectively fixed on the bracelet at positions corresponding to the conductive lugs on the two sides, and caulking grooves are formed in the inner ends of the electromagnets and used for accommodating the corresponding conductive lugs, so that the conductive lugs are not exposed in a natural state, and the wearing comfort is improved. The pulse circuit and the electromagnet are started, the electromagnet provides a repulsive force for the permanent magnet after conducting electricity, the repulsive force can force the inner elastic sheets to pop up inwards, after the inner elastic sheets at two ends pop up simultaneously, the copper sheet conductors at two sides can be enabled to contact muscle cortex on the arm respectively, the pulse circuit sends out pulse current, and the arm is stimulated to arouse a sleeping driver. After the driver is awake, radar equipment judges that it is in the situation state after detecting driver action signal, to camera acquisition element or come from the corresponding stop signal of cell-phone end transmission, and bracelet processor no longer provides impulse current.

Drawings

FIG. 1 is a block diagram of the system of the present invention.

Fig. 2 is a flow chart of the system of the present invention.

Fig. 3 is a side view of an autonomous radar apparatus.

Fig. 4 is a partial cross-sectional structural view of fig. 3.

Fig. 5 is a schematic cross-sectional structure of fig. 3.

Fig. 6 is a perspective view of fig. 3.

FIG. 7 is a schematic view of the gimbal adjustment block of FIG. 6.

FIG. 8 is a schematic view of another gimbal mount.

FIG. 9 is a side view of a universal screw shaft.

FIG. 10 is an assembled schematic view of a universal screw shaft.

Fig. 11 is a view showing the structure of a stationary base.

Fig. 12 is a view of the adjustable angle base.

Fig. 13 is a cassette structure view.

Fig. 14 is an electric pulse bracelet structure diagram.

Fig. 15 is a control block diagram of fig. 14.

Reference numbers in the figures: the radar device comprises a base 1, a universal adjusting seat 2, a filter cover 3, a radar device 4, a first plate layer 5, a second plate layer 6, a third plate layer 7, a left-right elastic sheet 8, a front-back elastic sheet 9, a left-right adjusting motor 10, a front-back adjusting motor 11, a universal joint 12, a screw 13, a shaft section 14, a support 15, a telescopic filter cover 16, an axial limiting block 17, an axial clamping groove 18, an adjusting sleeve 19, a mounting seat 20, a movable clamping seat 21, a sleeve 22, a disc spring 23, a static clamping seat 24, a circuit board 25, a signal wire 26, a connecting seat 27, a base 28, an adjusting arm 29, a lock pin 30, a support 31, a clamping ring 32, a lock wire 33, an adhesive layer 34, a rotating shaft 35, a pin shaft 36, a bracelet 37, a power display screen module 38, an electromagnet 39, an inner elastic sheet 40, a conductive bump 41, a through hole 42, a threaded sleeve 43, a steering motor 44 and a bearing 45.

Detailed Description

The invention is further illustrated with reference to the following figures and examples.

Example 1: a vehicle safety alarm system based on face recognition is disclosed, as shown in figure 1, and the system mainly comprises a radar detection unit, a camera acquisition and image processing device, a Bluetooth module, a mobile phone or vehicle-mounted electronic communication device, an early warning module, a cloud server and the like. The problems that the image processing information amount is too large, the energy consumption and storage space requirements are high and the like due to the fact that the camera is mainly used for collecting the image in the existing fatigue driving equipment are mainly solved, and the problems that the processing performance of the equipment is required to be configured is high and the cost is high are solved. Because the fatigue state is a limit state and an accidental event, the real-time camera shooting and image processing are adopted, so that the equipment is in an invalid starting operation stage for most of time, resources are wasted, and the service life of the equipment is influenced. The embodiment also aims at the problems that the existing fatigue driving early warning equipment has strong sensitivity and is easy to bring special interference to the driver, and the like.

The radar detection unit in the system comprises a support, an independent power supply or a vehicle-mounted power supply, radar equipment and a wireless transmission module such as Bluetooth. Radar equipment passes through the support to be fixed in the car in suitable position to with radar probe part regional contrast driver's cabin, one set of radar equipment can realize this function, does not exclude the condition of using a plurality of radar equipment, for example can adopt preceding radar and the cooperation of side radar. The radar detection unit and the camera acquisition unit establish a signal transmission relation in a wired or wireless mode, and because the radar detection unit only needs to provide the start opportunity for the camera acquisition unit, the communication is preferably established through the bluetooth, so that the radar detection unit and the camera acquisition unit can be respectively arranged at different positions in the vehicle. Unlike camera acquisition units that must be used to acquire facial features of the driver, radar detection units may be placed, for example, on the right side, front side, rear side, top, left side, or in hidden areas of the driver. Therefore, the arrangement modes of the radar detection units are more flexible and diversified.

The camera acquisition unit comprises a support, a monocular or binocular camera, a Bluetooth module, a processor and the like. For processing and monitoring the acquired images, but only for wirelessly transmitting the special images.

It should be noted that the camera acquisition unit necessity 1: for the collected driver state image data, not only the image processing but also the large data learning are required, and since the collected image data is data acquired under the necessary conditions, the degree of matching of the image data features with the drowsiness features is high, and the amount of image data is small and concentrated, so the processing amount is small. Necessity 2: the system is characterized by comprising a large number of transmission devices under a WIFI environment, only characteristic images are sent during travel, data transmission flow is reduced, feedback speed is increased, the system needs to establish a relationship with a cloud server platform, deep learning and training of labeled data are carried out through the cloud server platform, a training model is fed back to the vehicle-mounted device, the vehicle-mounted device utilizes an early-stage training model as a driver image data processing comparison basis under a WIFI-free environment, deep learning of specific drivers including facial features and the like is carried out by utilizing the cloud server platform under the WIFI environment, and labeling of the image data features is carried out through background technicians, so that reinforcement learning is achieved.

Regarding the supports of the radar detection unit and the camera acquisition unit, any form of fixing support or adjusting support can be adopted, and the embodiment is not described in detail.

The mobile phone or the vehicle-mounted mobile electronic equipment is characterized in that the camera acquisition unit sends image information to the APP end of the mobile phone or the vehicle-mounted mobile electronic equipment through the Bluetooth to store corresponding image data, and the APP end of the mobile phone or the vehicle-mounted mobile electronic equipment is connected with the cloud server platform to upload big data and download a training model. And transmitting the corresponding model data to a camera acquisition unit through a mobile phone or a vehicle-mounted mobile electronic device so as to facilitate detection and judgment of the vehicle running state.

The warning device can warn through a mobile phone, or warn a driver in a vibration or electric pulse mode through a special cushion or a special bracelet.

The radar is used for early warning monitoring, a photographing program is started to photograph when conditions are met, photographed images are transmitted with mobile phone Bluetooth, the mobile phone sends image information to a background cloud server to perform big data analysis, analysis results are transmitted to the mobile phone in real time, and the mobile phone sends out an alarm signal.

Regarding the image processing of the camera acquisition unit, a video is shot for the current state of a driver through the camera acquisition unit, the video is decoded into a plurality of pictures through an OpenCV code based on Python, and the method comprises the following steps of 4: the ratio of 1 is divided into a training set and a test set for a certain drowsiness of the driver, wherein 80% of the training set for a certain drowsiness of the driver is stored and 20% of the test set for a certain drowsiness of the driver is stored for judgment and detection. And manually labeling the drowsiness image data to be detected on the data set by the data labeling software LabelImg according to the historical cloud server platform. Training a data set on a Yolov4-tiny model by using a GPU on 2080Ti, pruning redundant channels and weight parameters in the trained Yolo-V4tiny drowsiness image data model by using a channel pruning algorithm, and applying the obtained weight parameters to the Yolov4-tiny model to detect the face recognition and drowsiness degree.

The method comprises the steps that a background provides a limited typical model of experience E for intensive learning and random acquisition, or acquires a large amount of unlabelled image data at a special moment, the image data is uploaded to a cloud platform in a network environment, the cloud platform automatically labels a newly uploaded image according to a large data sample artificially labeled by other images, and the newly uploaded image is downloaded to a vehicle-mounted terminal to be used as a model sample for on-site direct comparison. The camera acquisition unit applies the obtained model sample weight parameters to a Yolov4-tiny model to detect a rectangular frame of the face area of the driver in the real-time image and display the color of the frame as white. And then, carrying out image binarization on the real-time image after the data is labeled by setting a threshold value to be white corresponding to a gray value by utilizing the characteristic of image binarization so as to reserve a white rectangular frame of the area to be detected, filtering out redundant backgrounds except white so as to reduce the subsequent detection of redundant angular points, and finally detecting four angular points by utilizing a Harris algorithm of rectangular vertex characteristics to obtain pixel coordinates. And (4) combining the optimized three-dimensional calibration parameters and the pixel coordinates of the image to be detected with an SGBM three-dimensional matching algorithm to judge drowsiness so as to obtain a drowsiness image conclusion.

The apparatus performs image capturing and image processing work only under necessary conditions. Most of the time, the driver is in a non-fatigue state, and only the radar equipment with low energy consumption and low running cost is used for monitoring operation. The conditions for triggering the camera to start work and image processing can be controlled by setting sensitivity, but the radar is mainly used for detecting the human motion state, namely the radar acquires the head and neck motion state, the arm motion state, the body motion state and the like of a driver (other passengers) so as to determine that the driver is in a waking state. Only when the driver is in the stationary state t1 and when the vehicle is driven for a total of t2, the camera acquires necessary state information such as whether or not to close the eyes, whether or not to talk, etc., by capturing the detailed information of the driver's face, and processes the image to determine whether or not the driver is in a tired or sleeping state. The above is the selection characteristic of the intervention time of the camera acquisition unit, so that the relevance of the image information of the camera acquisition unit after intervention or driving and the drowsiness characteristic is strong, a large amount of irrelevant image information is eliminated, and the information amount and the cost of image processing are greatly simplified.

For the collected driver state image data with high association degree, big data learning needs to be participated in further, and the feature data which is preliminarily screened by the radar module and has high association degree with drowsiness can simplify the marking difficulty of big data features and improve the data training precision.

The system avoids the defects of low information processing speed, delayed feedback result and the like caused by real-time background communication, and can remarkably reduce data transmission flow and improve feedback speed by acquiring the training model and carrying out image processing under the condition of not giving background direct communication, thereby realizing the effect of light artificial intelligent image processing. The timing of establishing the relation between the system and the cloud server platform can be in any other non-driving state, deep learning and labeled data training are carried out through the cloud server platform, a training model is fed back to the vehicle-mounted equipment, and image data features are labeled through background technicians, so that learning is strengthened.

Example 2: an automatic control radar device suitable for a vehicle safety alarm system mainly comprises a base 1, a universal adjusting seat 2, a filter cover 3, a radar device 4, a display device and the like. The device can automatically adjust the radar capture range according to the collected radar signals, so that accurate driver action information is obtained, and interference caused by other action information is avoided. It can be seen that the radar device 4 is mainly used for acquiring the action change information of the driver, and the radar device is particularly sensitive to capturing the information, as shown in fig. 2, when the action information captured by the radar device is captured, the camera acquisition unit is triggered in parallel, but when no action information exists within t1, the radar device triggers the camera acquisition unit to start to work in cooperation. The method is characterized in that the ordinary radar equipment is easy to interfere when acquiring the action information, therefore, when the ordinary radar equipment is adopted in the system, interference factors need to be eliminated, particularly in the environment in a car with a small space, when the radar equipment captures the action information of a driver, the direction of a probe of the radar equipment is firstly adjusted to enable the radar equipment to face the driver, then, the area outside the driver is shielded through a filter cover 3, signals are captured only in the range of the driver, and finally, the driver is tracked through self-adaptive adjustment of a universal adjusting seat 2 according to the body position of the driver, so that the radar equipment is only used for automatically tracking and detecting the area near the driver.

Specifically, the universal adjusting base 2 includes a first plate layer 5, a second plate layer 6 and a third plate layer 7, as shown in fig. 4 and 7, a left-right spring 8 is fixed at the right end between the first plate layer 5 and the second plate layer 6, and a front-back spring 9 is fixed at the rear end between the second plate layer 6 and the third plate layer 7. Therefore, the first board layer 5 and the second board layer 6 can rotate left and right relatively, the second board layer 6 and the third board layer 7 can rotate back and forth relatively, and the combination of left and right rotation and back and forth rotation enables the first board layer 5 and the third board layer 7 to rotate back and forth and left and right to a proper position within a certain range.

As can be seen from fig. 3 and 4, a left-right adjusting motor 10 is fixed at the outer end of the first slab 5, and an adjusting screw is connected to a rotating shaft of the motor and is connected to a corresponding screw hole on the second slab 6. The outer end of the first plate layer 5 is also fixed with a front and back adjusting motor 11, a rotating shaft of the motor is connected with an adjusting screw, and the adjusting screw penetrates through a through hole 42 formed in the second plate layer 6 and then is connected into a corresponding screw hole in the third plate layer 7. The left-right adjusting motor 10 and the front-back adjusting motor 11 are controlled to rotate respectively, the first plate layer 5 and the second plate layer 6 can rotate left and right relatively, the second plate layer 6 and the third plate layer 7 rotate front and back relatively, the controller automatically controls the rotation degree of the two motors according to a radar device feedback signal, and then the first plate layer 5 and the third plate layer 7 can rotate front and back and left and right to a proper position within a certain range under automatic control.

As shown in fig. 5, the radar device 4 is installed in the center of the third floor 7, and the filter housing 3 is fitted around the outside of the radar device 4.

It can be seen from above structure and control relation that this embodiment radar equipment can independent operation, and the controller is according to radar equipment real-time feedback signal, and control regulation motor 10 is rotated and front and back regulation motor 11 is rotated about controlling for radar equipment's orientation can be along with driver's position and self-adaptation change and location, thereby can obtain only driver's action information.

Example 3: in addition to embodiment 2, universal screw shafts are respectively mounted on the rotating shafts of the left-right adjusting motor 10 and the front-back adjusting motor 11. The structure is shown in fig. 9 or fig. 10, the left-right adjusting motor 10 and the front-back adjusting motor 11 can be self-adaptive to bend along with the rotation of the corresponding plate layer, but can still transmit torsion, so that the screw 13 at the tail end of the torsion rotates. As can be seen in fig. 9, the universal screw shaft comprises a shaft at the head end, 14, a universal joint 12 in the middle and a screw 13 at the tail end. One combination is shown in fig. 10, a special-shaped hole such as a hexagonal clamping groove is formed in the end portion of a universal joint 12, a special-shaped column such as a hexagonal column is arranged at the end portion of a screw, the special-shaped column and the screw are matched and sleeved together, and the transmission distance of the screw 13 can be changed by replacing the screws with different lengths.

When a universal screw shaft is applied to the universal adjustment base 2, as shown in fig. 4, in the present embodiment, this structure can provide more rotation functions of the first sheet layer 5, the second sheet layer 6, and the third sheet layer 7, so that the rotation range can be increased relative to embodiment 2, so that the radar detection device can be placed closer to the driver.

Example 4: on the basis of the embodiment 2 or 3, the filter cover 3 is arranged to be of an axially telescopic adjusting structure. As shown in fig. 5, one adjustment method is that a support 15 is fixed at the outer end of the third plate layer 7, the support 15 includes a fixed sleeve, a telescopic filter cover 16 is sleeved in the fixed sleeve, an axial limit block 17 is arranged at the inner side of the fixed sleeve, a thread is arranged at the outer side of the telescopic filter cover 16, an axial slot 18 is recessed along the axial direction on the thread, and the axial limit block 17 is embedded in the axial slot 18. The outer side of the fixed sleeve is sleeved with a rotatable adjusting sleeve 19, the adjusting sleeve 19 is sleeved with the outer side of the telescopic filter cover 16, the adjusting sleeve 19 is in threaded connection with the telescopic filter cover 16, and the adjusting sleeve 19 can be rotated to drive the telescopic filter cover 16 to stretch axially.

A perforation is provided in the center of the third plate layer 7 and the radar device 4 is mounted. Specifically, considering that the radar apparatus has transmitting and receiving ends, the present embodiment can change the angular positions of transmission and reception by the structural form as shown in fig. 5. If a movable clamping seat 21 is arranged on a rear end mounting seat 20 of the radar transmitting and receiving probe, a static clamping seat 24 is fixed at the position of the through hole, a saw-toothed convex-concave connection structure is arranged on a butt joint surface between the movable clamping seat 21 and the static clamping seat 24, and a disc spring 23 is arranged after a sleeve 22 connected with the center of the movable clamping seat 21 penetrates through the through hole, so that the movable clamping seat 21 and the static clamping seat 24 have an elastic compression relationship, and the angular position of the radar equipment can be changed by rotating the radar equipment. The power supply and signal wires of the radar equipment are led out from the sleeve 22 and then connected with the circuit board 25, and the circuit board 25 is fixed between the board layers or in the base in the implementation.

Example 5: based on the above embodiments, the base is fixed on the connecting seat, and the connecting seat is fixed at any appropriate position in the vehicle body. Fig. 11, 12 and 13 respectively provide three kinds of fixed connection structures, as shown in fig. 11, the base 1 is fixed on the connecting seat 27, and the base 28 integrated with the connecting seat is fixed at a position lower than the vehicle body in the front windshield by adhering or attracting, so as to prevent the sight from being influenced as much as possible. As shown in fig. 12, on the basis of fig. 11, the connecting section 27 is connected with the base 28 through the rotating shaft 35, so that the connecting section 27 can rotate relative to the base 28, and meanwhile, the adjusting arms 29 are respectively hinged between the connecting section 27 and the base 28 through the pin shafts, and the ends of the two adjusting arms 29 are fixed through the lock pins 30, so that the connecting section 27 has an inclination angle relative to the base 28. As shown in figure 13, the device is fixed on the air outlet grille position of the air outlet of the air conditioner by adopting a clamping way, or is sleeved on the support position of the rearview mirror, as shown in the figure, a structure form is that a support 31 is fixed on a connecting seat, a clamping ring 32 is installed on the support, and locking wires 33 penetrate through the end parts of the clamping rings 32 at two sides, and the structure can be fixed on a rod piece below the headrest of the auxiliary seat backrest. And combinations of the above structures. And the antiskid mat is matched with the adhesive layer and placed on the instrument desk at a flat position close to the position under the front stop.

Example 6: on the basis of each above embodiment, a structural style of the warning device is as shown in fig. 14, the warning device is a bracelet type, and includes a bracelet 37, a power display screen module 38, an electromagnet 39 and an inner elastic sheet 40, the bracelet 37 is connected with the power display screen module 38, and under a natural state, the inner elastic sheet 40 is fixed at the bottom of the power display screen module 38 and is hidden in a groove of the bracelet 37. It can be seen from the figure that the two sides of the inner spring 40 are symmetrical, the two outer ends of the inner spring 40 are respectively provided with a conductive bump 41, the front side of the conductive bump 41 is a copper sheet conductor, the rear side is a permanent magnet, the two are combined together and fixed at the end part of the inner spring, and a power line connected with the copper sheet conductor is wrapped on the inner spring and led into the output end of the pulse circuit to be connected. The bracelet corresponding to the conductive bumps 41 on the two sides is respectively fixed with an electromagnet 39, and the inner end of the electromagnet is provided with an embedding groove for arranging the corresponding conductive bump 41, so that the conductive bump is not exposed in a natural state, and the wearing comfort is improved.

As can be seen from fig. 15, the power supply, the processor and the wireless receiving module are arranged in the power supply display screen module 38, the processor receives the warning signal from the camera acquisition unit or from the mobile phone end, the pulse circuit and the electromagnet are started, the electromagnet provides a repulsive force for the permanent magnet after being conductive, the repulsive force can force the inner spring to pop out inwards, after the inner springs at the two ends pop out simultaneously, the copper sheet conductors at the two sides can respectively contact the muscle cortex on the arm, and the pulse circuit sends out pulse current to stimulate the arm to further awaken a sleeping driver. After the driver is awake, radar equipment judges that it is in the situation state after detecting driver action signal, to camera acquisition element or come from the corresponding stop signal of cell-phone end transmission, and bracelet processor no longer provides impulse current.

It is to be understood that the above-described embodiments of the present invention are merely illustrative of or explaining the principles of the invention and are not to be construed as limiting the invention. Therefore, any modification, equivalent replacement, improvement and the like made without departing from the spirit and scope of the present invention should be included in the protection scope of the present invention. It is not excluded that after the through-holes are provided in the respective sheet layers, threaded sleeves 43 as shown in fig. 4 are fastened in the through-holes, into which threaded sleeves the screws of the universal screw shaft are screwed. As shown in fig. 8, another structure of the gimbal adjustment base 2 according to embodiment 2 includes only the first plate layer 5 and the second plate layer 6, and the right and left resilient pieces 8 are fixed to the right end portion between the first plate layer and the second plate layer, and a steering motor 44 is mounted to the center of the second plate layer 6 through a bearing 45, and the steering motor 44 is fixed to the base 1. And a left and right adjusting motor 10 is installed on the first sheet layer 5, the left and right adjusting motor 10 for automatically controlling the rotation between the first sheet layer 5 and the second sheet layer 6. Thus, when the first and second laminas 5 and 6 are rotated, the steering motor 44 is controlled to rotate, and the orientation of the radar apparatus can be changed.