阅读说明：本技术 基于人形目标识别的多雷达协同探测报警系统 (Multi-radar cooperative detection alarm system based on humanoid target recognition ) 是由 刘宏江 王军锋 丁国富 温忠凯 徐海军 王光辉 杨明宇 吕大勇 祁晓临 于 2021-09-24 设计创作，主要内容包括：本发明公开了基于人形目标识别的多雷达协同探测报警系统,包括：多个雷达探测设备均设置在铁路沿线的支杆上,每个雷达探测设备均检测进入铁路沿线的入侵目标并输出第一入侵信号；分别与每个雷达探测设备连接的控制器基于来自于每个雷达探测设备的第一入侵信号获得第二入侵信号；根据第二入侵信号采集入侵目标的图像的图像采集设备；与图像采集设备连接的主控平台基于来自于图像采集设备的图像进行人形目标识别,根据识别结果输出报警信号。本发明基于人形目标识别的多雷达协同探测报警系统通过多雷达协同探测,实现探测区域全覆盖,并通过人形目标定位方法进行铁路侵限行为控制和判断,对非法人员侵限的捕捉及判断更精准,降低系统的误报率。(The invention discloses a multi-radar collaborative detection alarm system based on humanoid target recognition, which comprises: the system comprises a plurality of radar detection devices, a first sensor, a second sensor, a third sensor and a fourth sensor, wherein the radar detection devices are all arranged on a supporting rod along the railway, and each radar detection device detects an intrusion target entering the railway and outputs a first intrusion signal; the controller respectively connected with each radar detection device obtains a second intrusion signal based on the first intrusion signal from each radar detection device; the image acquisition equipment acquires an image of the intrusion target according to the second intrusion signal; and the main control platform connected with the image acquisition equipment identifies the humanoid target based on the image from the image acquisition equipment and outputs an alarm signal according to an identification result. The multi-radar cooperative detection alarm system based on humanoid target recognition realizes full coverage of a detection area through multi-radar cooperative detection, controls and judges the railway invasion limit behavior through the humanoid target positioning method, captures and judges illegal personnel invasion limits more accurately, and reduces the false alarm rate of the system.)

1. A multi-radar cooperative detection alarm system based on humanoid target recognition is characterized by comprising:

the system comprises a plurality of radar detection devices, a first sensor, a second sensor and a controller, wherein each radar detection device is arranged on a support rod along the railway, and detects an intrusion target entering the railway and outputs a first intrusion signal;

the controller is respectively connected with each radar detection device, and obtains a second intrusion signal based on the first intrusion signal from each radar detection device and outputs the second intrusion signal;

the image acquisition equipment acquires an image of the intrusion target according to the second intrusion signal;

the main control platform is connected with the image acquisition equipment, and the main control platform identifies a humanoid target based on the image from the image acquisition equipment and outputs an alarm signal according to an identification result.

2. The system of claim 1, wherein the master platform performs humanoid target recognition based on the image from the image capture device comprises:

matching the image with a preset image to obtain a scene of the image;

segmenting the image into a plurality of regions based on a scene of the image;

respectively calculating the area value of each region;

respectively judging whether each area is consistent with a preset area;

selecting a region with an area value larger than an area threshold value and a region inconsistent with the preset region to generate a target region set;

and identifying the humanoid target based on the target area set and the humanoid target identification template.

3. The system of claim 2, wherein the human-shaped target recognition template comprises a plurality of human-shaped recognition sub-templates, and the human-shaped target recognition based on the target area set and the human-shaped target recognition template comprises:

calculating the pixel value of each human figure recognition sub-template;

and respectively calculating the pixel value of each region in the target region set, acquiring the difference value between the pixel value of each region and the pixel value of each human figure recognition sub-template, and determining that the intrusion target is a human when the difference value is greater than a pixel threshold value.

4. The system according to claim 3, wherein the human-shaped recognition sub-template comprises a square head area and a trapezoid shoulder area, and the head area is connected with the upper bottom of the shoulder area;

the area of the figure formed by the figures in the figure identifier template is as follows:

Area(hs)＝L_head*W_head+1/2(L1_shoulder+L2_shoulder)*H_shoulder

wherein L is_headAnd W_headLength and width of the head region, respectively, L1_shoulderAnd L2_shoulderUpper and lower bottoms, respectively, of shoulder regions, H_shoulderThe height of the shoulder region.

5. The system of claim 1, wherein the controller obtaining a second intrusion signal based on the first intrusion signal from each radar detection device comprises:

respectively carrying out filtering and denoising on a first intrusion signal of each radar detection unit to obtain a point trace after primary processing;

matching the point tracks after the primary processing in different scanning periods and different detection distances with the flight tracks in the database to obtain the motion tracks of the intrusion targets detected by the radar detection unit;

and fusing the motion tracks of the intrusion targets detected by the plurality of radar detection units to obtain a second intrusion signal.

6. The system of claim 5, wherein the fusing the motion tracks of the intrusion targets detected by the radar detection units to obtain the second intrusion signal comprises:

analyzing the motion trail of the intrusion target detected by each radar detection unit, and judging whether the motion trail of the intrusion target detected by the plurality of radar detection units is repeated according to the direction, angle and position of the intrusion target;

and deleting the repeated motion tracks of the invading targets, combining the motion tracks of the remaining invading targets and obtaining a second invading signal.

7. The system according to claim 6, wherein the detection orientations of the radar detection devices are different, and the detection orientation coverage of the radar detection devices is 360 °.

8. The system according to claim 1, further comprising:

the alarm device is connected with the main control platform, and the alarm system executes alarm after receiving the alarm signal;

the solar power supply equipment is respectively connected with the radar detection equipment, the controller and the image acquisition equipment, and the solar power supply equipment provides power for the radar detection equipment, the controller and the image acquisition equipment.

9. The system according to claim 1, wherein the image capturing device is a red-exposure-free dome camera.

10. The system according to claim 1, wherein the radar detection device is a frequency modulation continuous wave radar.

Technical Field

The invention belongs to the technical field of railway security alarm, and particularly relates to a multi-radar cooperative detection alarm system based on humanoid target recognition.

Background

The railway is the aorta of national economy, and plays an important role in the aspects of guaranteeing the stable operation of the national economy, maintaining the economic safety, the social stability and the like. According to statistics, the business mileage of China railways reaches over 13.9 kilometers (wherein the high-speed rail is 3.5 kilometers) by the end of 2019, and the China railways are the first to live in the world. Along with the rapid development of railway construction in China, the mileage of railway traffic is longer and longer, economic development and personnel traffic efficiency in various regions are greatly promoted, but due to the factors of wide regions, complex terrain structures, large climate change and the like in China, disaster accidents along the railway, particularly in mountainous areas and hilly lands frequently occur, such as rock rolling, debris flow and the like, the normal operation of the railway is blocked, on the other hand, a lot of potential safety hazards are brought, railway facilities are damaged, irrelevant personnel often happen regardless of the conditions that a railway protective net directly penetrates through the railway, and the like, so that the railway safety production is greatly damaged, and economic loss and personal injury are caused to a certain degree.

The intrusion detection alarm system with the railway operation characteristics is researched for safely, efficiently and conveniently ensuring the railway perimeter protection, and has great practical significance and economic benefit in the aspects of ensuring the railway safe operation management, improving the railway line safety protection capability, improving the accurate alarm of intrusion behaviors, effectively reducing the construction and maintenance cost and the like.

Therefore, a railway security alarm system which improves the railway line security protection capability and accurately alarms the intrusion behavior is particularly needed.

Disclosure of Invention

The invention aims to provide a multi-radar cooperative detection alarm system based on humanoid target recognition, which improves the safety protection capability of a railway line and accurately alarms intrusion behaviors.

In order to achieve the above object, the present invention provides a multi-radar cooperative detection alarm system based on humanoid target recognition, comprising: the system comprises a plurality of radar detection devices, a first sensor, a second sensor and a controller, wherein each radar detection device is arranged on a support rod along the railway, and detects an intrusion target entering the railway and outputs a first intrusion signal; the controller is respectively connected with each radar detection device, and obtains a second intrusion signal based on the first intrusion signal from each radar detection device and outputs the second intrusion signal; the image acquisition equipment acquires an image of the intrusion target according to the second intrusion signal; the main control platform is connected with the image acquisition equipment, and the main control platform identifies a humanoid target based on the image from the image acquisition equipment and outputs an alarm signal according to an identification result.

Optionally, the performing, by the master control platform, human-shaped target recognition based on the image from the image acquisition device includes: matching the image with a preset image to obtain a scene of the image; segmenting the image into a plurality of regions based on a scene of the image; respectively calculating the area value of each region; respectively judging whether each area is consistent with a preset area; selecting a region with an area value larger than an area threshold value and a region inconsistent with the preset region to generate a target region set; and identifying the humanoid target based on the target area set and the humanoid target identification template.

Optionally, the human-shaped target recognition template includes a plurality of human-shaped recognition sub-templates, and performing human-shaped target recognition based on the target area set and the human-shaped target recognition template includes: calculating the pixel value of each human figure recognition sub-template; and respectively calculating the pixel value of each region in the target region set, acquiring the difference value between the pixel value of each region and the pixel value of each human figure recognition sub-template, and determining that the intrusion target is a human when the difference value is greater than a pixel threshold value.

Optionally, the human-shaped identifier template includes a square head region and a trapezoidal shoulder region, and the head region is connected to the upper bottom of the shoulder region.

Optionally, the area of the figure formed by the figures in the figure recognition sub-template is as follows:

Area(hs)＝L_head*W_head+1/2(L1_shoulder+L2_shoulder)*H_shoulder

wherein L is_headAnd W_headLength and width of the head region, respectively, L1_shoulderAnd L2_shoulderUpper and lower bottoms, respectively, of shoulder regions, H_shoulderThe height of the shoulder region.

Optionally, the area of the figure formed by the figures in the figure recognition sub-template is as follows:

Area(hs)＝L_head*W_head+1/2(L1_shoulder+L2_shoulder)*H_shoulder

wherein L is_headAnd W_headLength and width of the head region, respectively, L1_shoulderAnd L2_shoulderUpper and lower bottoms, respectively, of shoulder regions, H_shoulderThe height of the shoulder region.

Optionally, the obtaining, by the controller, a second intrusion signal based on the first intrusion signal from each radar detection device includes: respectively carrying out filtering and denoising on a first intrusion signal of each radar detection unit to obtain a point trace after primary processing; matching the point tracks after the primary processing in different scanning periods and different detection distances with the flight tracks in the database to obtain the motion tracks of the intrusion targets detected by the radar detection unit; and fusing the motion tracks of the intrusion targets detected by the plurality of radar detection units to obtain a second intrusion signal.

Optionally, the fusing the motion trajectories of the intrusion targets detected by the plurality of radar detection units to obtain the second intrusion signal includes: analyzing the motion trail of the intrusion target detected by each radar detection unit, and judging whether the motion trail of the intrusion target detected by the plurality of radar detection units is repeated according to the direction, angle and position of the intrusion target; and deleting the repeated motion tracks of the invading targets, combining the deleted motion tracks of the invading targets and obtaining a second invading signal.

Optionally, the detection orientations of the plurality of radar detection devices are different, and the detection orientation coverage range of the plurality of radar detection devices is 360 °.

Optionally, the alarm system further includes: the alarm device is connected with the main control platform, and the alarm system executes alarm after receiving the alarm signal; the solar power supply equipment is respectively connected with the radar detection equipment, the controller and the image acquisition equipment, and the solar power supply equipment provides power for the radar detection equipment, the controller and the image acquisition equipment.

Optionally, the image acquisition device is a dome camera without red exposure.

Optionally, the radar detection device is a frequency modulation continuous wave system radar.

The invention has the beneficial effects that: the multi-radar cooperative detection alarm system based on humanoid target recognition realizes full coverage of a detection area through multi-radar cooperative detection, controls and judges the railway invasion limit behavior through the humanoid target recognition method, captures and judges illegal personnel invasion limits more accurately, reduces the false alarm rate of the system, improves the management efficiency of a railway management department, and improves the reliability of the system.

The present invention has other features and advantages which will be apparent from or are set forth in detail in the accompanying drawings and the following detailed description, which are incorporated herein, and which together serve to explain certain principles of the invention.

Drawings

The above and other objects, features and advantages of the present invention will become more apparent by describing in more detail exemplary embodiments thereof with reference to the attached drawings, in which like reference numerals generally represent like parts throughout.

Fig. 1 shows a block diagram of a multi-radar cooperative detection alarm system based on humanoid target recognition according to one embodiment of the present invention.

Fig. 2 shows a flow chart of humanoid target recognition performed by a master control platform of the multi-radar collaborative detection alarm system based on humanoid target recognition according to one embodiment of the invention.

FIG. 3 shows a multi-radar fusion processing flow diagram of the multi-radar cooperative detection alarm system based on humanoid target recognition according to one embodiment of the invention.

Fig. 4 shows a single radar coverage and angle diagram of a multi-radar cooperative detection alarm system based on humanoid target recognition according to an embodiment of the invention.

FIG. 5 shows a detailed block diagram of a multi-radar cooperative detection alarm system based on humanoid target recognition, according to one embodiment of the present invention.

Description of reference numerals:

102. a radar detection device; 104. a controller; 106. an image acquisition device; 108. and a master control platform.

Detailed Description

Preferred embodiments of the present invention will be described in more detail below. While the following describes preferred embodiments of the present invention, it should be understood that the present invention may be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

The invention discloses a multi-radar collaborative detection alarm system based on humanoid target recognition, which comprises: each radar detection device is arranged on a support rod along the railway, detects an intrusion target entering the railway and outputs a first intrusion signal; the controller is respectively connected with each radar detection device, and the controller obtains a second intrusion signal based on the first intrusion signal from each radar detection device and outputs the second intrusion signal; the image acquisition equipment acquires an image of the intrusion target according to the second intrusion signal; the main control platform is connected with the image acquisition equipment, carries out humanoid target recognition based on the image from the image acquisition equipment, and outputs an alarm signal according to a recognition result.

The multi-radar collaborative detection alarm system and device based on humanoid target recognition mainly comprise a main control platform, radar detection equipment, a controller, image acquisition equipment, other accessory equipment and the like.

1) Radar detection device

The radar detection equipment is the core composition of the system, and the equipment adopts an FMCW system radar with high precision and low power consumption to carry out covering detection on an intrusion area. The radar detection equipment can acquire information such as azimuth angles, moving speed and distance of the invading targets, and meanwhile, multi-target positioning and track tracking are supported. The system utilizes the characteristic of strong adaptability of the millimeter wave radar to the environment to realize all-weather monitoring along the railway all the day. The device has excellent detection effects in the daytime, at night, in rainy days and in haze days, combines advanced intelligent algorithm analysis, effectively filters interference of plants and small animals, and reduces the false alarm rate of the system.

2) Controller

The controller receives the information of each radar detection device, performs fusion processing on the signals of the radar detection devices to obtain a second intrusion signal, sends the second intrusion signal to the image acquisition device, and triggers the image acquisition device to acquire the image of the intrusion target according to the information of the second intrusion signal.

For the detection of the target area, the conventional approach usually adopts a single radar. The single radar detection area and the expansion angle are limited, the target area is difficult to cover by 360 degrees, and the multi-radar cooperative detection → data fusion → detection judgment → fusion output → image and video acquisition or monitoring calling → result confirmation is adopted, so that the function of one-monitoring multi-radar cooperative detection video confirmation of the target area is realized, and the system cost can be effectively reduced.

3) Image acquisition device

And after receiving a second intrusion signal of the controller, the image acquisition equipment acquires the image and video information of the intrusion target according to the second intrusion signal and transmits the acquired information to the master control platform so that the master control platform can recognize the human form.

4) Main control platform

Specifically, the Human Shape Recognition (HSR) method is a technology for finding, recognizing and positioning a Human Shape target in an imaging space by applying a certain processing technology to a graphic image by using certain characteristics of Human body imaging, and the Human Shape Recognition (HSR) technology is an important technology of an intelligent Security (SS) system, is a fusion of computer vision, pattern Recognition, an image processing technology and a morphology technology, and is an important research subject of the SS system. Many intelligent security and protection (SS) systems for detecting and positioning human forms are inevitably applied to human form recognition (HSR) technology, so that the human form recognition (HSR) technology has important practical application significance.

The invention uses a human-shaped target positioning method based on gray level characteristics to identify the intrusion target. The method is not only simple and easy to apply in some occasions with low processing speed, such as illegal invasion of personnel into railway lines, but also has high accuracy. Applications of other methods (a background difference method, an inter-frame difference method, an optical flow field method, a method based on a kalman filter, an inter-frame binary matching method, high-speed moving target detection based on matching fourier transform, a method based on three-frame difference and edge information, region-level moving target detection, moving target detection based on a dynamic scene image sequence, a method based on a codebook, a method based on a statistical background model, a method based on a probability statistics adaptive background model, and a method based on a 2D spatiotemporal entropy threshold) are often explored based on a certain purpose of research, although the accuracy of the algorithm may be improved to a certain extent, the complexity of the algorithm calculation is often higher, and thus the method is less used in practical applications.

Besides the human-shaped target recognition, the master control platform provides unified management of all monitoring areas in the jurisdiction range, and has the functions of real-time video viewing, accurate positioning, multi-source information management and the like. The accurate positioning is that after receiving the personnel intrusion alarm signal, the accident site can be immediately and accurately positioned; the multi-source information management is to uniformly summarize and manage multi-source information such as historical accident information and video data.

According to an exemplary embodiment, the multi-radar cooperative detection alarm system based on humanoid target recognition realizes the full coverage of a detection area through multi-radar cooperative detection, and emphasizes the condition of illegal limit intrusion of personnel in a prevention and control area through a humanoid target recognition method, so that the management efficiency of a railway management department is improved, and the false alarm rate of the system is reduced. The system is based on humanoid location and recognition algorithm to carry out railway and invade limit behavior control and judgement, consequently, the system invades the capture of limit and judges more accurately to illegal personnel, has reduced the system false alarm rate, promotes system's efficiency, in addition, still has following technological effect:

(1) and the robustness of the system and the device is enhanced. Because the radar detection data fusion technology is based on data collected by a plurality of radars, the influence of local environment change on the system can be minimized, and the system has stronger adaptability to the environment change.

(2) And the spatial coverage of the system is expanded. Because the area covered by a single radar is very limited, and the network formed by a plurality of radars greatly expands the search range of the radars, when a human-shaped target moves out of the scanning area of one radar, the human-shaped target may enter the scanning area of another radar, so that the target can be continuously tracked.

(3) Increasing the reliability of the system. The same target is detected by a plurality of radars, so that the error caused by radar detection can be reduced, and the position of the target can be determined more accurately.

(4) Improving the detection capability of the system. Under the condition of a certain false alarm density, the probability of finding a target by a multi-radar system is obviously higher than that of a single-radar system.

(5) The reliability of the system is improved. In the system, the failure of the working capacity of a single radar does not greatly affect the system, and due to the overlapping of coverage areas of the radars, other radars can make up the detection data lost due to the failure of the failed radar to a certain extent.

(6) The correctness of system decision is improved. Since the data relied upon by the railway manager to make decisions is provided by multiple radars, this data is more reliable than the data provided by a single radar, and therefore more helpful to the railway manager to make the correct decisions.

(7) The multi-radar detection fusion technology is taken as a leading factor, the man-shaped recognition and positioning algorithm technology is combined, the multi-radar and video cooperative linkage intelligent protection means is researched, the all-weather and three-dimensional target detection alarm capability is formed, and the system has the functions of illegal intrusion alarm, video tracking, voice drive-away and the like. When an illegal intrusion event occurs, the radar actively discovers and continuously tracks the target, the video linkage tracking is triggered, the video rechecks to confirm the authenticity of the intrusion target, and the voice calling equipment is automatically started to drive the target away from the line.

Alternatively, the main control platform performs human-shaped target recognition based on the image from the image acquisition device, and comprises: matching the image with a preset image to obtain a scene of the image; dividing the image into a plurality of regions based on a scene of the image; respectively calculating the area value of each region; respectively judging whether each area is consistent with a preset area; selecting a region with an area value larger than an area threshold value and a region inconsistent with a preset region to generate a target region set; and identifying the humanoid target based on the target area set and the humanoid target identification template.

Specifically, in order to select a target region as much as possible on an image and reduce the amount of calculation, 3 assumptions are given here, that (1) a foreground region with an excessively small area is unlikely to be a human body; (2) the bending of the human body in the horizontal direction should not be too large; (3) the human body does not appear outside the delineated non-target area.

In order to select the maximum target area, a target area segmentation method based on horizontal and vertical vector sets is proposed herein. The specific method is that a horizontal vector group A and a vertical vector group B are established first, and then horizontal and vertical search is carried out. In the horizontal search, a global window is established, full search is carried out in the vertical direction, if a pixel point exists in the corresponding vertical direction, the corresponding horizontal vector group is marked as 1, and if not, the corresponding horizontal vector group is marked as 0. Vertical searching is similar to horizontal direction, but requires fixing the horizontal direction and sliding the window in the vertical direction. It is seen that in the obtained horizontal vector group and vertical vector group, the area region is obtained when the number of consecutive 1 s and 0 s is more.

Based on the above assumptions, methods based on maximum area detection are proposed. Firstly, area-based region segmentation is carried out on a full scene, then the maximum region to be segmented is determined according to the area of the region and the number of pixels, a plurality of segmented regions are obtained, the area value of each region is respectively calculated, the region with the area value larger than the area threshold value and the region inconsistent with the preset region are selected and reserved, a target region set is generated, each region in the target region set is matched with a human-shaped target identification template, and human-shaped target identification is carried out.

As an alternative, the human-shaped target recognition template comprises a plurality of human-shaped recognition sub-templates, and the human-shaped target recognition based on the target area set and the human-shaped target recognition template comprises: calculating the pixel value of each human figure recognition sub-template; and respectively calculating the pixel value of the region aiming at each region in the target region set, acquiring the difference value between the pixel value of the region and the pixel value of each human figure recognition sub-template, and determining the invading target person when the difference value is greater than the pixel threshold value.

Specifically, a plurality of human figure recognition sub-templates are established, a pixel value of a region in a target region set is calculated according to the region, and a difference value between the pixel value of the region and the pixel value of each human figure recognition sub-template is obtained; and when the difference value between the pixel value of the area and the pixel value of a certain human figure recognition sub-template is greater than the pixel threshold value, determining that the intrusion target is a human.

Alternatively, the human shape recognition sub-template comprises a square head region and a trapezoidal shoulder region, and the head region is connected with the upper bottom of the shoulder region.

Alternatively, the area of the figure formed by the figures in the figure recognition sub-template is as follows:

Area(hs)＝L_head*W_head+1/2(L1_shoulder+L2_shoulder)*H_shoulder

wherein L is_headAnd W_headLength and width of the head region, respectively, L1_shoulderAnd L2_shoulderUpper and lower bottoms, respectively, of shoulder regions, H_shoulderThe height of the shoulder region.

Specifically, the image exhibits its visual features in the human visual range, including the shape, color, texture, shading, etc. of the image content. The visual characteristics are very visual and visual, so the visual characteristics are more easily accepted and utilized by people. The visual characteristics of human targets are numerous, and human-shaped target recognition based on shape characteristics is increasingly studied and applied by people. The extraction and expression of image features are important contents of target detection, and the image features refer to visual features (such as color, texture, shape, object surface and the like) of an image. The figure-of-the-human recognizer template presented herein includes a square head region and a trapezoidal shoulder region, the head region being connected to the upper base of the shoulder region.

The area composition of the region is shown as the following formula:

Area(hs)＝L_head*W_head+1/2(L1_shoulder+L2_shoulder)*H_shoulder

this template divides the foreground region into A, B two parts. Wherein A is the real head and shoulder area of the human body, B is the external limiting area of the template, and the establishment evaluation function is shown as the following formula:

Area＝1/Area(hs)(∑hs-∑b)

in the above formula, Σ hs and Σ b are respectively represented by the following formulas:

wherein L is_headAnd W_headLength and width of the head region, respectively, L1_shoulderAnd L2_shoulderThe upper and lower bottoms of the shoulder reminding templates, H_shoulderThe height of the shoulder trapezoidal template. In order to indicate a certain offset which may occur due to the position of the head and shoulders of the human body in the standing state, the variable α is used to indicate the relative offset of the head and shoulders. Area (hs) and area (b) represent the collective areas of the head-shoulder regions in the template, respectively. And Σ hs and Σ b denote the size of the foreground pixels in the head-shoulder region. As can be seen from the formula, when the obtained head-shoulder portion coincides with the real image imaging portion, area (hs) ═ hs, that is, the head-shoulder set area is equivalent to the number of pixels. From this, it is known that Area ∈ [0,1 ]]。

In practical image quantization, the area of the head region is often larger than the shoulder region, and in order to effectively indicate the area of the head and the shoulder region, the following formula is given:

Area(shoulder)＝βArea(head)

the proportional relation of the areas of the head area and the shoulder area is shown, the value is smaller than 0.5, if the value is overlarge, the areas of the head and the shoulder are equivalent, the turning part between the head and the shoulder is easy to detect weakly, a continuous area is easy to detect, and the precision is greatly reduced. If the value is too small, the area of the head region is large, the detection precision of the shoulder part is reduced, the situation that the head is a large region, but the shoulder region is only a small region is often caused, a sliding value range is given for effective calibration, and beta belongs to [0.3,0.5], and the change range can accurately calibrate the area regions of the head and the shoulder according to the difference of specific head and shoulder imaging in actual detection.

Alternatively, the controller obtaining the second intrusion signal based on the first intrusion signal from each radar detection device includes: respectively carrying out filtering and denoising on a first intrusion signal of each radar detection unit to obtain a point trace after primary processing; matching the point tracks after the primary processing in different scanning periods and different detection distances with the flight tracks in the database to obtain the motion tracks of the intrusion targets detected by the radar detection unit; and fusing the motion tracks of the intrusion targets detected by the plurality of radar detection units to obtain a second intrusion signal.

As an alternative, fusing the motion trajectories of the intrusion targets detected by the multiple radar detection units, and obtaining the second intrusion signal includes: analyzing the motion trail of the intrusion target detected by each radar detection unit, and judging whether the motion trail of the intrusion target detected by the plurality of radar detection units is repeated according to the direction, angle and position of the intrusion target; and deleting the repeated motion tracks of the invading targets, combining the deleted motion tracks of the invading targets and obtaining a second invading signal.

Specifically, the radar detection equipment is a core component of the system, and adopts an FMCW system radar with high precision and low power consumption to carry out covering detection on an intrusion area. The radar detection equipment can acquire information such as azimuth angles, positions, moving speeds and distances of the invading targets, and meanwhile, multi-target positioning and track tracking are supported. The controller analyzes the information of azimuth angle, position, moving speed, distance and the like of the intrusion target output by each radar detection device, judges whether repeated information exists or not, deletes repeated first intrusion signals and takes the remaining first intrusion signals as second intrusion signals. The alarm system utilizes the characteristic of strong adaptability of the millimeter wave radar to the environment to realize all-weather monitoring along the railway all the day.

The radar information cooperative detection and processing technology obtained by a plurality of radar detection devices is the core of a radar information processing system. The radar information processing technology is divided into three levels, namely radar signal processing and detection, single radar data processing and multi-radar data fusion, and is also summarized into radar information primary processing, radar information secondary processing and radar information tertiary processing.

1) Radar signal processing and target detection

In the radar information processing technology, radar signal processing and target detection are referred to as radar information primary processing. The radar information is processed once, usually in units of radar, which is to process raw data of a certain scanning period of the radar. The purpose of the primary processing of the radar information is to extract useful target information in a noisy background. Although a lot of filtering and noise reduction technologies are adopted in the processing process, the clutter cannot be completely filtered due to the limitation of the clutter wave spectrum characteristic and the filter performance, and the remaining clutter and interference signals are called clutter residual.

After one processing, if there are more clutter left, the burden of subsequent processing will be increased, and the computer will be saturated. Furthermore, the outliers and the target splits caused by the distance segmentation and target detection criteria also require merging here. The above work may be understood as a pre-treatment before the secondary treatment.

2) Single radar data processing

The single radar data processing is also called radar information secondary processing. In the secondary processing, the data processing unit takes the trace points output by the primary processing as input, associates the trace points with the trace points in the database, updates the state of the target, and simultaneously performs extrapolation, filtering and other processing, namely the tracking of the target. Here, the association means that the trace points of the target are paired with the corresponding trace points.

3) Multi-radar data fusion

The three-pass processing based on the results of the two-pass processing is typically performed at the fusion center of the multi-radar data fusion system. The processing of multiple radar tracks is generally called multi-radar data fusion, and the fusion center receives the point track data transmitted by each radar and performs subsequent data processing. The device has excellent detection effects in the daytime, at night, in rainy days and in haze days, combines advanced intelligent algorithm analysis, effectively filters interference of plants and small animals, and reduces the false alarm rate of the system.

Alternatively, the detection orientations of the plurality of radar detection devices are different, and the detection orientations of the plurality of radar detection devices cover 360 °.

Specifically, the full coverage of a detection area is realized through multi-radar cooperative detection, and the illegal limit invasion condition of personnel in the control area is mainly prevented and controlled through a humanoid target positioning method, so that the management efficiency of a railway management department is improved.

The main functions realized by the multiple radars in a cooperative way are as follows: the method comprises the steps of dividing an alarm area, automatically and continuously acquiring data of a protection area, discovering an invading target in real time, acquiring information such as an azimuth angle, a position and a moving speed of the invading target, performing linkage video rechecking, performing linkage sound-light alarm, generating a detection log, performing a fault self-checking function, performing self-networking and self-powering. The performance indexes are as follows: the real-time detection distance of a single radar is not less than 200 meters; the positioning precision of the invasion target is better than 1 meter; detecting an intrusion target by not less than 30cm multiplied by 30 cm; supporting multi-target simultaneous detection and alarm; the false alarm rate is less than or equal to 1 percent, and the missing report rate is less than or equal to 2 percent; the multi-radar cooperative detection and effective coverage range is 360 degrees.

As an alternative, the alarm system further comprises: the alarm device is connected with the main control platform, and the alarm system executes alarm after receiving the alarm signal; the solar energy power supply equipment is respectively connected with the plurality of radar detection equipment, the controller and the image acquisition equipment, and the solar energy power supply equipment provides power for the plurality of radar detection equipment, the controller and the image acquisition equipment.

Specifically, alarm device still includes other auxiliary assembly facilities, for example for a plurality of radar detection equipment, controller and image acquisition equipment provide the solar energy power supply unit of power, adopts the solar energy power supply to improve whole alarm system's life, and alarm device includes audible alarm, light warning, shouting to report to the police and pronunciation intercom system etc..

As an alternative, the image acquisition device is a red-exposure-free dome camera.

Specifically, the image acquisition equipment selects a high-definition red-exposure-free ball machine suitable for railway application, and is matched with radar detection equipment to realize continuous tracking of the intrusion target. The main functions realized are as follows: wide area monitoring, video rechecking alarm information, capturing an intrusion target picture and uploading, intercepting intrusion target video data and uploading, fault self-checking, self-networking and self-powering. The performance indexes that can be achieved are: the real-time detection distance is not less than 200 meters; the horizontal rotation of 360 degrees is supported, and the rotation of minus 20 degrees to 90 degrees is realized in the vertical direction.

Alternatively, the radar detection device is a frequency modulated continuous wave system radar.

Examples

Fig. 1 shows a block diagram of a multi-radar cooperative detection alarm system based on humanoid target recognition according to one embodiment of the present invention. Fig. 2 shows a flow chart of humanoid target recognition performed by a master control platform of the multi-radar collaborative detection alarm system based on humanoid target recognition according to one embodiment of the invention. FIG. 3 shows a multi-radar fusion processing flow diagram of the multi-radar cooperative detection alarm system based on humanoid target recognition according to one embodiment of the invention. Fig. 4 shows a single radar coverage and angle diagram of a multi-radar cooperative detection alarm system based on humanoid target recognition according to an embodiment of the invention. FIG. 5 shows a detailed block diagram of a multi-radar cooperative detection alarm system based on humanoid target recognition, according to one embodiment of the present invention.

Referring to fig. 1, fig. 2, fig. 3, fig. 4 and fig. 5, the multi-radar cooperative detection alarm system based on humanoid target recognition comprises: each radar detection device 102 is arranged on a strut along the railway, and each radar detection device 102 detects an intrusion target entering the railway and outputs a first intrusion signal; the controller 104 is connected with each radar detection device 102, and the controller 102 obtains a second intrusion signal based on the first intrusion signal from each radar detection device 102 and outputs the second intrusion signal; the image acquisition equipment 106 acquires an image of the intrusion target according to the second intrusion signal; and the main control platform 108 is connected with the image acquisition equipment 106, and the main control platform 108 identifies human-shaped targets based on the images from the image acquisition equipment 106 and outputs alarm signals according to the identification results.

The main control platform 108 performs human-shaped target recognition based on the image from the image acquisition device, including: matching the image with a preset image to obtain a scene of the image; dividing the image into a plurality of regions based on a scene of the image; respectively calculating the area value of each region; respectively judging whether each area is consistent with a preset area; selecting a region with an area value larger than an area threshold value and a region inconsistent with a preset region to generate a target region set; and identifying the humanoid target based on the target area set and the humanoid target identification template.

The humanoid target recognition template comprises a plurality of humanoid recognition sub-templates, and the humanoid target recognition is carried out based on the target area set and the humanoid target recognition template, and comprises the following steps: calculating the pixel value of each human figure recognition sub-template; and respectively calculating the pixel value of the region aiming at each region in the target region set, acquiring the difference value between the pixel value of the region and the pixel value of each human figure recognition sub-template, and determining the invading target person when the difference value is greater than the pixel threshold value.

The human shape recognition sub-template comprises a square head area and a trapezoidal shoulder area, and the head area is connected with the upper bottom of the shoulder area.

Wherein, the area of the figure formed by the figures in the figure identifier template is as follows:

Area(hs)＝L_head*W_head+1/2(L1_shoulder+L2_shoulder)*H_shoulder

wherein L is_headAnd W_headLength and width of the head region, respectively, L1_shoulderAnd L2_shoulderAre respectively asUpper and lower bottoms of shoulder regions, H_shoulderThe height of the shoulder region.

Wherein, the area of the figure formed by the figures in the figure identifier template is as follows:

Area(hs)＝L_head*W_head+1/2(L1_shoulder+L2_shoulder)*H_shoulder

wherein L is_headAnd W_headLength and width of the head region, respectively, L1_shoulderAnd L2_shoulderUpper and lower bottoms, respectively, of shoulder regions, H_shoulderThe height of the shoulder region.

Wherein the controller obtaining a second intrusion signal based on the first intrusion signal from each radar detection device comprises: respectively carrying out filtering and denoising on a first intrusion signal of each radar detection unit to obtain a point trace after primary processing; matching the point tracks after the primary processing in different scanning periods and different detection distances with the flight tracks in the database to obtain the motion tracks of the intrusion targets detected by the radar detection unit; and fusing the motion tracks of the intrusion targets detected by the plurality of radar detection units to obtain a second intrusion signal.

Wherein, fuse the motion trajectory of the invasion target that a plurality of radar detection units detected, obtain the second invasion signal and include: analyzing the motion trail of the intrusion target detected by each radar detection unit, and judging whether the motion trail of the intrusion target detected by the plurality of radar detection units is repeated according to the direction, angle and position of the intrusion target; and deleting the repeated motion tracks of the invading targets, combining the deleted motion tracks of the invading targets and obtaining a second invading signal.

Wherein, the detection position of a plurality of radar detection equipment is different, and the detection position coverage of a plurality of radar detection equipment is 360.

Wherein, alarm system still includes: the alarm device is connected with the main control platform, and the alarm system executes alarm after receiving the alarm signal; the solar energy power supply equipment is respectively connected with the plurality of radar detection equipment, the controller and the image acquisition equipment, and the solar energy power supply equipment provides power for the plurality of radar detection equipment, the controller and the image acquisition equipment.

Wherein the image acquisition equipment is a dome camera without red exposure.

Wherein, the radar detection equipment is a radar made of frequency modulation continuous wave.

The radar information cooperative detection and processing technology is the core of a radar information processing system. The radar information processing technology is divided into three levels, namely radar signal processing and detection, single radar data processing and multi-radar data fusion, and is also summarized into radar information primary processing, radar information secondary processing and radar information tertiary processing.

1) Radar signal processing and target detection

In the radar information processing technology, radar signal processing and target detection are referred to as radar information primary processing. The radar information is processed once, usually in units of radar, which is to process raw data of a certain scanning period of the radar. The purpose of the primary processing of the radar information is to extract useful target information in a noisy background. Although a lot of filtering and noise reduction technologies are adopted in the processing process, the clutter cannot be completely filtered due to the limitation of the clutter wave spectrum characteristic and the filter performance, and the remaining clutter and interference signals are called clutter residual.

After one processing, if there are more clutter left, the burden of subsequent processing will be increased, and the computer will be saturated. Furthermore, the outliers and the target splits caused by the distance segmentation and target detection criteria also require merging here. The above work may be understood as a pre-treatment before the secondary treatment.

2) Single radar data processing

The single radar data processing is also called radar information secondary processing. In the secondary processing, the data processing unit takes the trace points output by the primary processing as input, associates the trace points with the trace points in the database, updates the state of the target, and simultaneously performs extrapolation, filtering and other processing, namely the tracking of the target. Here, the association means that the trace points of the target are paired with the corresponding trace points.

3) Multi-radar data fusion

The three-pass processing based on the results of the two-pass processing is typically performed at the fusion center of the multi-radar data fusion system. The processing of multiple radar tracks is generally called multi-radar data fusion, and the fusion center receives the point track data transmitted by each radar and performs subsequent data processing. The flow of radar information processing is shown in fig. 3.

As shown in fig. 5, the alarm system further includes a PC terminal, a mobile phone terminal user, a data server, and the like, and fully combines the internet design concept, and takes a human-shaped positioning recognition algorithm as a starting point, and effectively integrates a multi-radar detection device, video monitoring, voice equipment, and the like by relying on advanced technologies such as artificial intelligence, edge calculation, mobile communication, and the like, so as to construct an intelligent and three-dimensional security system facing a railway key protection area. The front-end detection equipment is deployed on site, the background is deployed on a local Internet website group platform, and the PC end is placed in an on-duty room or a monitoring center.

Having described embodiments of the present invention, the foregoing description is intended to be exemplary, not exhaustive, and not limited to the embodiments disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments.