阅读说明：本技术 一种基于360度环视摄像头的车位尺寸识别系统及方法 (Parking space size identification system and method based on 360-degree all-round-looking camera ) 是由 马世典 江中旭 方伟锋 李玥 于 2019-09-18 设计创作，主要内容包括：本发明公开了一种基于360度环视摄像头的车位尺寸识别系统及方法,该系统包括中央控制器、图像采集模块、图像处理控制器、无线传输控制器和数据库。搭载该系统的智能汽车进行自动泊车时,通过360度环视摄像头拍摄空闲泊车位和停放车辆的泊车位的图像。经过图像处理控制器进行图像处理后,分别获取空闲泊车位的长宽比和车位中车辆与车位的长宽比,并通过图像处理控制器的识别匹配技术,将拍摄到该车辆的图像与数据库中的车辆信息分析匹配,智能识别出该车辆的品牌型号,获取该车辆的外部尺寸(长度与宽度),然后通过简单的比例换算从而得到泊车位的尺寸(宽度及深度)。该系统适用性强、识别率高,能大大提高泊车时的车位尺寸识别精度。(The invention discloses a parking space size recognition system and method based on a 360-degree all-around camera. When the intelligent automobile carrying the system is used for automatic parking, images of idle parking spaces and parking spaces for parking vehicles are shot through the 360-degree all-round camera. After image processing is carried out by the image processing controller, the length-width ratio of an idle parking space and the length-width ratio of a vehicle and a parking space in the parking space are respectively obtained, the shot image of the vehicle is analyzed and matched with vehicle information in the database through the identification matching technology of the image processing controller, the brand model of the vehicle is intelligently identified, the external size (length and width) of the vehicle is obtained, and then the size (width and depth) of the parking space is obtained through simple proportional conversion. The system has strong applicability and high recognition rate, and can greatly improve the recognition precision of the parking space size when parking.)

1. The invention discloses a parking space size recognition system based on a 360-degree all-round camera, which is characterized in that: the system comprises a central controller, an image acquisition module, an image processing controller and a wireless transmission controller, wherein the image acquisition module, the image processing controller and the wireless transmission controller are connected with the central controller;

the central controller is responsible for controlling and monitoring the operation of the image acquisition module and the image processing controller, and realizes the information transmission between the database and the image processing controller through the wireless transmission controller;

the image acquisition module is responsible for acquiring image information around the vehicle and comprises illumination equipment, an image acquisition controller, a 360-degree all-round-looking camera and an image memory which are sequentially connected; the lighting device comprises a light sensor and an LED lamp;

the system also comprises a database, wherein the database has a networking function, the central controller can control the database to update networking data through the wireless controller, and the database has complete images of various vehicles and model size information corresponding to the vehicles;

when an intelligent automobile equipped with the system comes to a parking lot or a parking space is searched near the parking space arranged regularly on the roadside, images of an idle parking space and a parking space where a vehicle is parked are shot through a 360-degree all-round camera, after the shot images are processed, the length-width ratio of the parking space and the length-width ratio of the vehicle and the parking space in the parking space are respectively obtained, meanwhile, through an image recognition technology, the image information of the vehicle in the shot parking space is analyzed and compared with the image information of the vehicle in a database, the brand and the model of the vehicle are recognized, the external size of the vehicle is obtained, and therefore the size of the parking space is obtained through proportional conversion.

2. The parking space size recognition system based on the 360-degree all-round camera as claimed in claim 1, wherein: the 360-degree all-round looking camera is a plurality of camera assemblies arranged on the front side, the rear side, the left side and the right side of the automobile, and is a wide-angle camera which can collect images of the surrounding environment of the automobile in real time.

3. The parking space size recognition system based on the 360-degree all-round camera as claimed in claims 1 and 2, wherein: the image acquisition module is responsible for the collection to vehicle information around, and wherein, image acquisition controller control 360 degrees look around the camera and shoot and begin and stop to image memory is given with the image or the image transmission of shooing.

4. The parking space size recognition system based on the 360-degree all-round camera as claimed in claim 1, wherein: when the lighting equipment is in dim light of a parking environment, such as a night parking environment or an environment with insufficient lighting of an underground parking lot, the light sensors send signals to the acquisition controller according to the intensity of light, the acquisition controller controls the on and off of the LEDs, and the light sensors and the LED lamps are arranged on the four sides of the vehicle body and can effectively assist the 360-degree all-round camera to work.

5. The parking space size recognition system based on the 360-degree all-round camera as claimed in claim 1, wherein: the image processing controller is responsible for processing and identifying image information, and distortion correction, aerial view transformation, image splicing, image preprocessing and image identification of the collected images are achieved through the image processing controller.

6. The parking space size recognition system based on the 360-degree all-round camera as claimed in claim 1, wherein: the wireless transmission controller is responsible for information interaction between the central controller and the database, and wireless data transmission is realized through a compiled communication protocol.

7. The parking space size recognition method based on the 360-degree all-round camera as claimed in claim 1, wherein: the method comprises the following steps:

step 1) when an intelligent automobile searches for a parking space, a driver starts the system, an optical sensor collects surrounding illumination intensity information and sends the information to an acquisition controller, and if the parking environment is dim in light, the acquisition controller controls LED lamps around an automobile body to be turned on to assist in illumination;

step 2) the acquisition controller controls the 360-degree all-around camera to shoot images of the surrounding environment of the vehicle, and the acquired image information is numbered and stored in the image memory;

step 3) the central controller controls the image processing controller to call image information from the image memory, and the image processing controller identifies and processes the image information, firstly identifies an image of a parking space area, and then further identifies an idle parking space image and a parking space image of a parked vehicle;

step 4) acquiring the length-width ratio of the parking space as S by processing the image of the idle parking space; acquiring a parking space image of a parked vehicle, wherein the ratio of the length of the parking space to the length of the vehicle is T1, or the ratio of the width of the parking space to the width of the vehicle is T2, and segmenting the vehicle image of the parking space;

step 5) the central controller controls the wireless transmission controller to send the vehicle images in the database to the image processing controller, the image processing controller matches the segmented vehicle images with the vehicle images in the database, and size information of the vehicle is obtained after the matching is successful, wherein the length of the vehicle is l, and the width of the vehicle is w;

step 6) the central controller calculates the length of the parking space as L ═ T1 × L, and the width is

8. The parking space size recognition method based on the 360-degree all-round camera as claimed in claim 7, wherein: in the step 3):

the method for identifying the idle parking space image comprises the following steps: the image processing controller extracts the pictures shot by the 360-degree all-round looking cameras from the image memory, and numbers and corresponds the pictures, namely, the pictures shot by the four cameras at the same time are a group, and then image processing is carried out, and the method comprises the following steps:

and 3.1, carrying out distortion correction on the images: calibrating the 360-degree around-looking camera by a Zhang Zhengyou calibration method to obtain internal parameters and external parameters of the camera, and then carrying out distortion correction on the image;

and 3.2, carrying out aerial view transformation processing on the image, wherein the aerial view transformation can be realized by an inverse perspective transformation algorithm, and a general transformation formula [ x ', y ', w ' ] of the perspective transformation is as follows:

where (u, v) are the original image pixel coordinates, for the transformed image pixel coordinates,

3.3, cutting and splicing the processed image to form a complete 360-degree all-round view image;

and 3.4, preprocessing the image, including filtering and denoising and image graying processing, eliminating sharp noise in the image by adopting a mean value filtering method, and obtaining an image g (x, y) after the image f (i, j) is subjected to smooth denoising through a mean value filtering algorithm, wherein the equation is as follows:

wherein M is the total number of pixels including the current pixel in the template, and i, j are pixel coordinates; the template operator is generally m × m;

the calculation formula of the image gray scale is as follows:

F(x，y)＝0.299×R(x，y)+0.587×G(x，y)+0.114×B(x，y)

after all pixel points in the color image are converted by the above formula, the R (x, y), G (x, y) and B (x, y) color images are converted into gray images;

and 3.5, identifying a rectangular area in the image, wherein the specific process is as follows:

carrying out canny edge detection on the image, extracting the outline of the image, identifying a rectangle through Hough transformation, extracting the outline of a rectangular region, calculating the area of the outline, and leaving out the rectangular region which does not accord with the area of the parking space on the way;

and 3.6, identifying the parking space angle, which specifically comprises the following steps:

four points of a rectangular area, namely coordinates of parking space angles, are obtained by searching for the convex hulls, and the coordinates of the four parking space angles in the rectangular area accord with the length-width ratio of the parking spaces, so that the parking spaces are considered to be effective parking spaces, and the length-width ratio of the parking spaces is S;

the method for identifying the parking space image of the parked vehicle comprises the following steps: the image processing controller copies the pictures extracted from the image memory, the pictures are respectively an A group picture and a B group picture, the A group picture and the B group picture are respectively numbered, and then the image processing is carried out, the image processing method specifically comprises the following steps:

group A images:

carrying out distortion correction on the image according to the obtained internal and external parameters of the camera;

preprocessing the image, including filtering, denoising and eliminating sharp noise in the image by means of image mean value, and enhancing the details of the image by gamma conversion;

editing positive and negative samples of the vehicle, training an SVM classifier for recognizing the vehicle, recognizing the vehicle through the SVM classifier, and taking the region as a region of interest;

extracting the outline of the region of interest through canny edge detection, and extracting the vehicle outline through adjusting a threshold value;

editing positive and negative samples of the parking space angle, training an SVM classifier for recognizing the parking space angle, and recognizing the parking space angle through the SVM classifier;

determining whether the vehicle is in a parking space area or not according to the area of the parking space angle, and if so, segmenting the vehicle in the parking space from the surrounding background through semantic segmentation;

identifying and matching the segmented vehicle image with a vehicle image in a database by a neural network method to obtain the model of the vehicle, and then obtaining the size information of the vehicle to obtain the length l and the width w of the vehicle;

group B images:

extracting the images of the vehicles identified in the group A corresponding to the serial number in the group B;

carrying out distortion correction on the image according to the obtained internal and external parameters of the camera;

and editing positive and negative samples of the parking space angle, training SVM classifiers for identifying front and rear car lights, identifying the front and rear car lights in the car image, and marking the outer sides of the car lights. If the image is two front lamps or two rear lamps, the image is the front side or the rear side of the vehicle body, and if the image is the front lamps and the rear lamps, the image is the left side or the right side of the vehicle body;

identifying and marking the position angle through a classifier of the SVM for identifying the position angle;

performing bird's-eye view conversion processing on the image;

calculating the length proportion of the two parking space angles to the outer sides of the two lamps, and if the image is the front side or the rear side of the vehicle body, taking the length proportion of the parking space line to the outer sides of the two lamps as the proportion T2 of the parking space width to the vehicle width; if the image is the left side or the right side of the vehicle body, the length ratio of the parking space line to the outer sides of the two lamps is used as the ratio T1 of the parking space length to the vehicle length.

Technical Field

The invention belongs to the field of automatic parking, and particularly relates to a parking space size identification system and method based on vision.

Background

With the development of cities, city land use planning is more rigorous, parking spaces are narrower, efficient parking schemes are sought, fatigue of drivers is relieved, automatic parking systems are born, wherein the parking systems based on ultrasonic radars are mainstream technical schemes in parking space identification technologies, but due to the fact that ultrasonic radars are large in scattering angle and poor in directivity, when targets at a longer distance are measured, echo signals of the ultrasonic radars are weak, measurement accuracy is affected, and accurate size information of the parking spaces cannot be obtained.

In addition, can also realize the parking stall discernment through camera technique, can realize the discernment of car position line through the parking stall photo of on-vehicle camera shooting, but can not acquire the size of parking stall.

Disclosure of Invention

Aiming at the problem that the parking space size cannot be accurately obtained by the ultrasonic radar and the camera in the above situation, the invention provides a parking space size identification system and method based on a 360-degree look-around camera, so as to improve the accuracy of parking space size identification.

In order to achieve the purpose, the specific technical scheme of the invention is as follows:

a parking space size recognition system based on a 360-degree all-round-looking camera comprises a central controller, an image acquisition module, an image processing controller and a wireless transmission controller, wherein the image acquisition module, the image processing controller and the wireless transmission controller are connected with the central controller;

the central controller is responsible for controlling and monitoring the operation of the image acquisition module and the image processing controller, and realizes the information transmission between the database and the image processing controller through the wireless transmission controller;

the image acquisition module is responsible for acquiring image information around the vehicle and comprises illumination equipment, an image acquisition controller, a 360-degree all-round-looking camera and an image memory which are sequentially connected; the lighting device comprises a light sensor and an LED lamp;

the system also comprises a database, wherein the database has a networking function, the central controller can control the database to update networking data through the wireless controller, and the database has complete images of various vehicles and model size information corresponding to the vehicles;

when an intelligent automobile equipped with the system comes to a parking lot or a parking space is searched near the parking space arranged regularly on the roadside, images of an idle parking space and a parking space where a vehicle is parked are shot through a 360-degree all-round camera, after the shot images are processed, the length-width ratio of the parking space and the length-width ratio of the vehicle and the parking space in the parking space are respectively obtained, meanwhile, through an image recognition technology, the image information of the vehicle in the shot parking space is analyzed and compared with the image information of the vehicle in a database, the brand and the model of the vehicle are recognized, the external size of the vehicle is obtained, and therefore the size of the parking space is obtained through proportional conversion.

Furthermore, the 360-degree all-round looking camera is a plurality of camera assemblies arranged on the front side, the rear side, the left side and the right side of the automobile, and the camera is a wide-angle camera and can acquire images of the surrounding environment of the automobile in real time.

Further, the image acquisition module is responsible for the collection to vehicle information around, and wherein, the image acquisition controller control 360 degrees look around the camera and shoot and start and stop to image memory is given with the image or the image transmission who shoots.

Further, when the light of the parking environment is dim, such as a night parking environment or an environment with insufficient illumination of an underground parking lot, the light sensor sends a signal to the acquisition controller according to the intensity of the light, and the acquisition controller controls the on and off of the LED. The optical sensor and the LED lamps are arranged on four sides of the vehicle body, and the 360-degree all-round-looking camera can be effectively assisted to work.

Furthermore, the image processing controller is responsible for processing and identifying image information, and distortion correction, aerial view transformation, image splicing, image preprocessing and image identification of the collected images are realized through the image processing controller.

Further, the wireless transmission controller is responsible for information interaction between the central controller and the database, and wireless data transmission is achieved through the compiled communication protocol.

The invention discloses a parking space size identification method based on a 360-degree all-round camera, which adopts the technical scheme that the method comprises the following steps:

step 1) when an intelligent automobile searches for a parking space, a driver starts the system, an optical sensor collects surrounding illumination intensity information and sends the information to an acquisition controller, and if the parking environment is dim in light, the acquisition controller controls LED lamps around an automobile body to be turned on to assist in illumination;

step 2) the acquisition controller controls the 360-degree all-around camera to shoot images of the surrounding environment of the vehicle, and the acquired image information is numbered and stored in the image memory;

step 3) the central controller controls the image processing controller to call image information from the image memory, and the image processing controller identifies and processes the image information, firstly identifies an image of a parking space area, and then further identifies an idle parking space image and a parking space image of a parked vehicle;

step 4) acquiring the length-width ratio of the parking space as S by processing the image of the idle parking space; acquiring a parking space image of a parked vehicle, wherein the ratio of the length of the parking space to the length of the vehicle is T1, or the ratio of the width of the parking space to the width of the vehicle is T2, and segmenting the vehicle image of the parking space;

step 5) the central controller controls the wireless transmission controller to send the vehicle images in the database to the image processing controller, the image processing controller matches the segmented vehicle images with the vehicle images in the database, and size information of the vehicle is obtained after the matching is successful, wherein the length of the vehicle is l, and the width of the vehicle is w;

step 6) the central controller calculates the length of the parking space as L ═ T1 × L, and the width is

Or the parking space length is T2 xw xS, and the width is W2 xw.

Further, in the step 3):

the method for identifying the idle parking space image comprises the following steps: the image processing controller extracts the pictures shot by the 360-degree all-round looking cameras from the image memory, and numbers and corresponds the pictures, namely, the pictures shot by the four cameras at the same time are a group, and then image processing is carried out, and the method comprises the following steps:

and 3.1, carrying out distortion correction on the images: calibrating the 360-degree around-looking camera by a Zhang Zhengyou calibration method to obtain internal parameters and external parameters of the camera, and then carrying out distortion correction on the image;

and 3.2, carrying out aerial view transformation processing on the image, wherein the aerial view transformation can be realized by an inverse perspective transformation algorithm, and a general transformation formula [ x ', y ', w ' ] of the perspective transformation is as follows:

where (u, v) are the original image pixel coordinates,

for the transformed image pixel coordinates,

is a perspective transformation matrix;

3.3, cutting and splicing the processed image to form a complete 360-degree all-round view image;

and 3.4, preprocessing the image, including filtering and denoising and image graying processing, eliminating sharp noise in the image by adopting a mean value filtering method, and obtaining an image g (x, y) after the image f (i, j) is subjected to smooth denoising through a mean value filtering algorithm, wherein the equation is as follows:

wherein M is the total number of pixels including the current pixel in the template, and i, j are pixel coordinates; the template operator is generally m × m;

the calculation formula of the image gray scale is as follows:

F(x，y)＝0.299×R(x，y)+0.587×G(x，y)+0.114×B(x，y)

after all pixel points in the color image are converted by the above formula, the R (x, y), G (x, y) and B (x, y) color images are converted into gray images;

and 3.5, identifying a rectangular area in the image, wherein the specific process is as follows:

carrying out canny edge detection on the image, extracting the outline of the image, identifying a rectangle through Hough transformation, extracting the outline of a rectangular region, calculating the area of the outline, and leaving out the rectangular region which does not accord with the area of the parking space on the way;

and 3.6, identifying the parking space angle, which specifically comprises the following steps:

four points of a rectangular area, namely coordinates of parking space angles, are obtained by searching for the convex hulls, and the coordinates of the four parking space angles in the rectangular area accord with the length-width ratio of the parking spaces, so that the parking spaces are considered to be effective parking spaces, and the length-width ratio of the parking spaces is S;

the method for identifying the parking space image of the parked vehicle comprises the following steps: the image processing controller copies the pictures extracted from the image memory, the pictures are respectively an A group picture and a B group picture, the A group picture and the B group picture are respectively numbered, and then the image processing is carried out, the image processing method specifically comprises the following steps:

group A images:

carrying out distortion correction on the image according to the obtained internal and external parameters of the camera;

preprocessing the image, including filtering, denoising and eliminating sharp noise in the image by means of image mean value, and enhancing the details of the image by gamma conversion;

editing positive and negative samples of the vehicle, training an SVM classifier for recognizing the vehicle, recognizing the vehicle through the SVM classifier, and taking the region as a region of interest;

extracting the outline of the region of interest through canny edge detection, and extracting the vehicle outline through adjusting a threshold value;

editing positive and negative samples of the parking space angle, training an SVM classifier for recognizing the parking space angle, and recognizing the parking space angle through the SVM classifier;

determining whether the vehicle is in a parking space area or not according to the area of the parking space angle, and if so, segmenting the vehicle in the parking space from the surrounding background through semantic segmentation;

identifying and matching the segmented vehicle image with a vehicle image in a database by a neural network method to obtain the model of the vehicle, and then obtaining the size information of the vehicle to obtain the length l and the width w of the vehicle;

group B images:

extracting the images of the vehicles identified in the group A corresponding to the serial number in the group B;

carrying out distortion correction on the image according to the obtained internal and external parameters of the camera;

and editing positive and negative samples of the parking space angle, training SVM classifiers for identifying front and rear car lights, identifying the front and rear car lights in the car image, and marking the outer sides of the car lights. If the image is two front lamps or two rear lamps, the image is the front side or the rear side of the vehicle body, and if the image is the front lamps and the rear lamps, the image is the left side or the right side of the vehicle body;

identifying and marking the position angle through a classifier of the SVM for identifying the position angle;

performing bird's-eye view conversion processing on the image;

calculating the length proportion of the two parking space angles to the outer sides of the two lamps, and if the image is the front side or the rear side of the vehicle body, taking the length proportion of the parking space line to the outer sides of the two lamps as the proportion T2 of the parking space width to the vehicle width; if the image is the left side or the right side of the vehicle body, the length ratio of the parking space line to the outer sides of the two lamps is used as the ratio T1 of the parking space length to the vehicle length.

The invention has the following technical effects: the system has strong applicability and high recognition rate, and can greatly improve the recognition precision of the parking space size when parking. After image processing is carried out by the image processing controller, the length-width ratio of an idle parking space and the length-width ratio of a vehicle and a parking space in the parking space are respectively obtained, the shot image of the vehicle is analyzed and matched with vehicle information in the database through the identification matching technology of the image processing controller, the brand model of the vehicle is intelligently identified, the external size (length and width) of the vehicle is obtained, and then the size (width and depth) of the parking space is obtained through simple proportional conversion.

In addition, the identification method of the invention can not only obtain the accurate size information of the parking space; and can also realize the parking stall through 360 degrees look around camera techniques and discern, can realize the discernment of car position line through the parking stall picture that on-vehicle camera was shot, and acquire the size of parking stall easily.

Drawings

Fig. 1 is a schematic structural diagram of a parking space size recognition system based on a 360-degree all-round camera according to the present invention.

Fig. 2 is a flowchart of a parking space size identification method based on a 360-degree all-round camera according to the present invention.

FIG. 3 is a flow chart of an idle parking space image processing and recognition algorithm of the parking space size recognition system and method based on the 360-degree all-round camera of the present invention.

FIG. 4 is a flowchart of a parking space image processing and recognition algorithm for a parked vehicle based on a parking space size recognition system and method with a 360-degree look-around camera according to the present invention.

Detailed Description

The invention is further described below with reference to the accompanying drawings:

the parking space size identification system based on the 360-degree all-around camera is structurally shown in fig. 1, and the system component comprises a central controller, an image acquisition module, an image processing controller, a wireless transmission controller and a database. The image acquisition module comprises an image acquisition controller, a 360-degree all-round-looking camera, an image memory and lighting equipment; the modules and the controller are all modules and controllers which are commonly used in the market and are well-known technical modules which are easily obtained in the field.

The central controller is responsible for controlling and monitoring the operation of the image acquisition module and the image processing controller, and realizes the information transmission between the database and the image processing controller through the wireless transmission controller.

The image acquisition module is responsible for the collection to vehicle information around, and wherein, acquisition controller control 360 degrees look around the camera and shoot and begin and stop to image memory is given with the image or the image transmission of shooing. The lighting equipment comprises a light sensor and an LED lamp, the light sensor is responsible for sensing the light intensity around the vehicle, when the light does not meet the shooting condition of the 360-degree panoramic camera, a signal is sent to the acquisition controller, the LED lamp is turned on by the acquisition controller, and the auxiliary lighting is used for enabling the panoramic camera to normally work.

The 360-degree all-round looking camera is four cameras arranged on the front side, the rear side, the left side and the right side of the automobile, and the number of the cameras can be increased according to actual conditions. 360 degree look around the camera can acquire the environmental information of all directions alone, also can obtain 360 degrees panorama information all around of vehicle through the image concatenation, and the mounted position of camera is as follows: the front camera is arranged on the middle net, the rear camera is arranged at the handle of the trunk, and the front camera and the rear camera are positioned on the middle axis, so that the heights of pictures shot in front and at the rear are consistent; the left camera and the right camera are arranged below the left rear-view mirror and the right rear-view mirror, and the left camera and the right camera are consistent in height of pictures shot left and right. The camera adopts a fisheye lens, the visual angle is larger than 180 degrees so as to be convenient for splicing and displaying images, and the camera has enough definition so as to carry out the next image processing and recognition work.

When the light of the parking environment of the lighting device is dim, such as the night parking environment or the environment with insufficient lighting of an underground parking lot, the light sensor sends a signal to the acquisition controller according to the intensity of the light, and the acquisition controller controls the on and off of the LED. The optical sensor and the LED lamps are arranged on four sides of the vehicle body, and the 360-degree all-round-looking camera can be effectively assisted to work.

The image processing controller is responsible for processing and identifying image information, and distortion correction, aerial view transformation, image splicing, image preprocessing and image identification of the collected images are achieved through the image processing controller.

The wireless transmission controller is responsible for information interaction between the central controller and the database, and wireless data transmission is realized through a compiled communication protocol.

The database has a networking function, the central controller can control the database to update networking data through the wireless controller, and the database has complete images of various vehicles and model size information corresponding to the vehicles.

The parking space size identification method based on the 360-degree all-round camera is shown in fig. 2, when an intelligent automobile equipped with the system comes to a parking lot or starts to search parking spaces near the parking spaces with orderly arrangement on roadsides, images of idle parking spaces and parking spaces where vehicles are parked are shot through the 360-degree all-round camera, after the shot images are processed, the length-width ratio of the parking spaces and the length-width ratio of the vehicles and the parking spaces in the parking spaces are respectively obtained, meanwhile, through an image identification technology, the image information of the vehicles in the shot parking spaces is analyzed and compared with the image information of the vehicles in a database, the brand of the vehicle is identified, the external shape size (length and width) of the vehicle is obtained, and therefore the size (width and depth) of the parking spaces is obtained through scaling.

The method specifically comprises the following steps:

(1) when the intelligent automobile equipped with the system is used for searching for a parking space, a driver starts the system, and the light sensor acquires ambient light intensity information and sends the ambient light intensity information to the acquisition controller. If the parking environment is dim in light, the acquisition controller controls the LED lamps around the vehicle body to be turned on to assist in lighting.

(2) The acquisition controller controls the 360-degree all-round-looking camera to shoot images of the surrounding environment of the vehicle, and the acquired image information is numbered and stored in the image memory.

(3) The central controller controls the image processing controller to call image information from the image memory, and the image processing controller identifies and processes the image information, firstly identifies an image of a parking space area, and then further identifies an image of a free parking space and an image of a parking space of a parked vehicle.

(4) Acquiring the length-width ratio of the parking space as S by processing the image of the idle parking space; the ratio of the parking space length to the vehicle length is T1 (or the ratio of the parking space width to the vehicle width is T2) through the parking space image of the parked vehicle, and the vehicle image of the parking space is segmented.

(5) The central controller controls the wireless transmission controller to send the vehicle images in the database to the image processing controller, the image processing controller matches the divided vehicle images with the vehicle images in the database, size information of the vehicle is obtained after matching is successful, and the length l of the vehicle is w.

(6) Central controller computingThe length of the parking space is L ═ T1 × L, and the width is

(or the parking space length is T2 xw xS, and the width is W2 xw).

The flow chart of the idle parking space image processing and identifying algorithm of the parking space size identifying system and method based on the 360-degree all-round cameras is shown in fig. 3, and the image processing controller extracts the pictures shot by the 360-degree all-round cameras from the image memory and numbers the pictures correspondingly, namely, the pictures shot by the four cameras at the same time are in a group. Then, image processing is carried out, and the method specifically comprises the following steps:

1. these images are subjected to distortion correction. Calibrating the 360-degree around-looking camera by a Zhang Zhengyou calibration method to obtain internal parameters and external parameters of the camera, and then carrying out distortion correction on the image;

2. carrying out aerial view transformation processing on the image, wherein the aerial view transformation can be realized by an inverse perspective transformation algorithm, and the general transformation formula of the perspective transformation is as follows:

where (u, v) are the original image pixel coordinates,

for the transformed image pixel coordinates,

is a perspective transformation matrix.

3. Cutting and splicing the processed images to form a complete 360-degree all-round view image;

4. the image is preprocessed, the preprocessing comprises the steps of filtering denoising and image graying processing, a mean value filtering method is adopted in the image to eliminate sharp noise in the image, the image f (i, j) is subjected to smooth denoising through a mean value filtering algorithm to obtain an image g (x, y), and the equation is as follows:

where M is the total number of pixels in the template including the current pixel, and the template operator is generally M × M.

The calculation formula of the image gray scale is as follows:

F(x，y)＝0.299×R(x，y)+0.587×G(x，y)+0.114×B(x，y)

after all the pixel points in the color image are converted by the above formula, the color image is converted into a gray image;

5. identifying a rectangular area in the image, which comprises the following specific steps:

carrying out canny edge detection on the image, extracting the outline of the image, identifying a rectangle through Hough transformation, extracting the outline of a rectangular region, calculating the area of the outline, and leaving out the rectangular region which does not accord with the area of the parking space on the way;

6. the method for identifying the parking space angle specifically comprises the following steps:

four points of the rectangular area, namely coordinates of the parking space corners, are obtained by searching the convex hulls, the coordinates of the four parking space corners in the rectangular area accord with the length-width ratio of the parking space, the parking space is considered to be an effective parking space, and the length-width ratio of the parking space is S.

The parking space image processing and identifying algorithm flow chart of the parked vehicle based on the parking space size identifying system and method of the 360-degree all-round camera is shown in fig. 4, pictures extracted from an image memory by an image processing controller are copied, are respectively an A group picture and a B group picture, are respectively numbered, and are then processed, and the parking space image processing method specifically comprises the following steps:

group A images:

1. carrying out distortion correction on the image according to the obtained internal and external parameters of the camera;

2. preprocessing the image, including filtering, denoising and eliminating sharp noise in the image by means of image mean value, and enhancing the details of the image by gamma conversion;

3. editing positive and negative samples of the vehicle, training an SVM classifier for recognizing the vehicle, recognizing the vehicle through the SVM classifier, and taking the region as a region of interest;

4. extracting the outline of the region of interest through canny edge detection, and extracting the vehicle outline through adjusting a threshold value;

5. editing positive and negative samples of the parking space angle, training an SVM classifier for recognizing the parking space angle, and recognizing the parking space angle through the SVM classifier;

6. determining whether the vehicle is in a parking space area or not according to the area of the parking space angle, and if so, segmenting the vehicle in the parking space from the surrounding background through semantic segmentation;

7. and identifying and matching the segmented vehicle image with the vehicle image in the database by a neural network method to obtain the model of the vehicle, and then obtaining the size information of the vehicle to obtain the length l and the width w of the vehicle.

Group B images:

1. extracting the images of the vehicles identified in the group A corresponding to the serial number in the group B;

2. carrying out distortion correction on the image according to the obtained internal and external parameters of the camera;

3. and editing positive and negative samples of the parking space angle, training SVM classifiers for identifying front and rear car lights, identifying the front and rear car lights in the car image, and marking the outer sides of the car lights. If the image is two front lamps or two rear lamps, the image is the front side or the rear side of the vehicle body, and if the image is the front lamps and the rear lamps, the image is the left side or the right side of the vehicle body;

4. identifying and marking the position angle through a classifier of the SVM for identifying the position angle;

4. performing bird's-eye view conversion processing on the image;

5. calculating the length proportion of the two parking space angles to the outer sides of the two lamps, and if the image is the front side or the rear side of the vehicle body, taking the length proportion of the parking space line to the outer sides of the two lamps as the proportion T2 of the parking space width to the vehicle width; if the image is the left side or the right side of the vehicle body, the length ratio of the parking space line to the outer sides of the two lamps is used as the ratio T1 of the parking space length to the vehicle length.

In the above embodiment, the SVM classifier may be implemented by directly calling a self-contained SVM algorithm general module through MATLAB.

In the description herein, references to the description of the term "one embodiment," "some embodiments," "an illustrative embodiment," "an example," "a specific example," or "some examples" or the like mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

While embodiments of the invention have been shown and described, it will be understood by those of ordinary skill in the art that: various changes, modifications, substitutions and alterations can be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.