Blind Spot Pedestrian Detection Method and System

Abstract

The present invention provides a blind spot pedestrian detection method and system. The method includes: detecting a road surface image collected by a target camera to obtain the position of a pedestrian; inputting the road surface image into a preset semantic segmentation network to obtain a road surface segmentation map, and determining different blind spots according to the road surface segmentation map; determining the blind spot where the pedestrian is located according to the position of the pedestrian, and giving an alarm prompt to pedestrians in different blind spots; wherein, the target camera is installed in the driving vision blind spot. The system executes the method. The present invention can accurately and real-time detect the dangerous areas in the driving vision blind spot, and give an alarm prompt to pedestrians when they enter the dangerous areas of the blind spot, which can reduce the occurrence of safety accidents.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular, to a blind - area pedestrian detection method and system. Background Art

[0002] As is well known, due to the design of the vehicle itself, there are many blind areas that cannot be reached by the rear - view mirrors around the vehicle body. When pedestrians enter the blind area of the driving vision, there are great potential safety hazards, especially at the moment when the vehicle turns.

[0003] Currently, monitoring algorithms for blind areas have been gradually popularized on vehicles. However, the area covered by the camera's perspective is larger than the blind area. Moreover, for example, pedestrians outside the green belt and isolation fence are actually not in the dangerous area. If the dangerous area in the perspective is not restricted, frequent blind - area pedestrian alarms will instead affect the driver. Summary of the Invention

[0004] The blind - area pedestrian detection method and system provided by the present invention are used to solve at least one of the above problems existing in the prior art. It can accurately and real - time detect the dangerous area in the blind area of the driving vision, and give an alarm prompt to pedestrians when they enter the dangerous area of the blind area, which can reduce the occurrence of safety accidents.

[0005] A blind - area pedestrian detection method provided by the present invention includes:

[0006] Detect the road surface image collected by the target camera to obtain the position of the pedestrian;

[0007] Input the road surface image into a preset semantic segmentation network to obtain a road surface segmentation map, and determine different blind areas according to the road surface segmentation map;

[0008] Determine the blind area where the pedestrian is located according to the position of the pedestrian, and give an alarm prompt to pedestrians in different blind areas;

[0009] Wherein, the target camera is installed in the blind area of the driving vision.

[0010] According to the blind - area pedestrian detection method provided by the present invention, the detecting the road surface image collected by the target camera to obtain the position of the pedestrian includes:

[0011] Detect the road surface image based on the object - detection yolov5 algorithm to obtain the position of the pedestrian.

[0012] According to the blind - area pedestrian detection method provided by the present invention, the determining different blind areas according to the road surface segmentation map includes:

[0013] Perform linear fitting on the road surface segmentation map to obtain a road - edge fitting line;

[0014] Determine different blind spots according to the curb fitting line and the warning lines calibrated at different positions of the vehicle body in the preset target camera.

[0015] According to a blind spot pedestrian detection method provided by the present invention, the step of linearly fitting the road surface segmentation map to obtain a curb fitting line includes:

[0016] Perform denoising processing on the road surface segmentation map based on erosion and dilation processing;

[0017] Cluster the target pixel points in the denoised road surface segmentation map according to a preset spatial clustering algorithm;

[0018] Determine a target cluster containing the most of the target pixel points according to the clustering result;

[0019] If the ratio of the number of target pixel points in the target cluster to the total number of pixel points in the road surface segmentation map is greater than or equal to a preset value, then perform linear fitting on the target pixel points in the target cluster to obtain the curb fitting line.

[0020] According to a blind spot pedestrian detection method provided by the present invention, the step of determining the blind spot where the pedestrian is located according to the position of the pedestrian and giving an alarm prompt to pedestrians in different blind spots includes:

[0021] If the position of the pedestrian is within the range of the area near the vehicle body and the first warning line, determine that the blind spot where the pedestrian is located is the first blind spot, and give a first-level alarm prompt to the pedestrians in the first blind spot;

[0022] If the position of the pedestrian is within the range of the area near the vehicle body and the second warning line, determine that the blind spot where the pedestrian is located is the second blind spot, and give a second-level alarm prompt to the pedestrians in the second blind spot;

[0023] If the position of the pedestrian is within the range of the area near the vehicle body and the third warning line, determine that the blind spot where the pedestrian is located is the third blind spot, and give a third-level alarm prompt to the pedestrians in the third blind spot;

[0024] Among them, the area near the vehicle body is determined according to the area formed between the curb fitting line and the vehicle body;

[0025] The first warning line is determined according to the warning line at the first position of the vehicle body;

[0026] The second warning line is determined according to the warning line at the second position of the vehicle body;

[0027] The third warning line is determined according to the warning line at the third position of the vehicle body.

[0028] According to a blind - zone pedestrian detection method provided by the present invention, after determining the blind zone where a pedestrian is located according to the position of the pedestrian and giving warning prompts to pedestrians in different blind zones, it further includes:

[0029] According to the warning level of the warning prompt, send different first prompt instructions to the multimedia intelligent central control screen installed on the vehicle:

[0030] If the warning prompt is the first - level warning prompt, send a first prompt instruction to repeat the first target voice for a first preset number of times to the multimedia intelligent central control screen;

[0031] If the warning prompt is the second - level warning prompt, send a first prompt instruction to repeat the first target voice for a second preset number of times to the multimedia intelligent central control screen;

[0032] If the warning prompt is the third - level warning prompt, send a first prompt instruction to repeat the first target voice for a third preset number of times to the multimedia intelligent central control screen.

[0033] According to a blind - zone pedestrian detection method provided by the present invention, after determining the blind zone where a pedestrian is located according to the position of the pedestrian and giving warning prompts to pedestrians in different blind zones, it further includes:

[0034] According to the warning level of the warning prompt, send different second prompt instructions to the sound - light alarm installed on the vehicle:

[0035] If the warning prompt is the first - level warning prompt, send a second prompt instruction to turn on the light flashing and repeat the second target voice to the sound - light alarm;

[0036] If the warning prompt is the second - level warning prompt, send a second prompt instruction to turn on the light flashing and repeat the third target voice to the sound - light alarm;

[0037] If the warning prompt is the third - level warning prompt, send a second prompt instruction to turn on the light flashing and repeat the fourth target voice to the sound - light alarm.

[0038] The present invention also provides a blind - zone pedestrian detection system, including: a pedestrian detection module, a blind - zone determination module, and a warning prompt module;

[0039] The pedestrian detection module is used to detect the road surface image collected by the target camera to obtain the position of the pedestrian;

[0040] The blind - zone determination module is used to input the road surface image into a preset semantic segmentation network to obtain a road surface segmentation map, and determine different blind zones according to the road surface segmentation map;

[0041] The warning prompt module is used to determine the blind area where the pedestrian is located according to the position of the pedestrian, and give warning prompts to pedestrians in different blind areas;

[0042] Among them, the target camera is installed in the blind area of the driving vision.

[0043] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the steps of the blind area pedestrian detection method as described in any one of the above are implemented.

[0044] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the blind area pedestrian detection method as described in any one of the above are implemented.

[0045] The blind area pedestrian detection method and system provided by the present invention can accurately and real-time detect the dangerous area in the blind area of the driving vision, and give an alarm prompt to the pedestrian when the pedestrian enters the dangerous area of the blind area, which can reduce the occurrence of safety accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to these drawings.

[0047] Figure 1 is a schematic flowchart of the blind area pedestrian detection method provided by the present invention;

[0048] Figure 2 is a schematic diagram of the application scenario of the blind area pedestrian detection method provided by the present invention;

[0049] Figure 3 is a schematic structural diagram of the blind area pedestrian detection system provided by the present invention;

[0050] Figure 4 is a schematic physical structure diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0051] In order to make the purpose, technical solutions and advantages of the present invention clearer, the following will clearly and completely describe the technical solutions in the present invention in conjunction with the drawings in the present invention. Obviously, the described embodiments are some embodiments of the present invention, rather than all embodiments. Based on the embodiments in the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the protection scope of the present invention.

[0052] Figure 1 FIG. 1 is a flow chart of the blind spot pedestrian detection method provided by the present invention, as shown in FIG. Figure 1 As shown, the method includes:

[0053] S1. Detect the road image captured by the target camera to obtain the location of pedestrians;

[0054] S2. Inputting the road surface image into a preset semantic segmentation network to obtain a road surface segmentation map, and determining different blind spots based on the road surface segmentation map;

[0055] S3. Determine the pedestrian's blind spot based on the pedestrian's position and issue warnings to pedestrians in different blind spots;

[0056] Among them, the target camera is installed in the blind spot of driving vision.

[0057] It should be noted that the execution subject of the above method can be a computer device or a vehicle-mounted terminal. The following takes the vehicle-mounted terminal as an example to implement the blind spot pedestrian detection method provided by the present invention to explain the present invention in detail:

[0058] Alternatively, as Figure 2 As shown, the target camera sends the road image collected during driving to the vehicle terminal. The vehicle terminal receives the road image collected by the target camera and detects the received road image to locate the position of the pedestrian in the road image, thereby obtaining the position of the pedestrian.

[0059] The target camera can be a camera with infrared function. This is mainly because blind spot monitoring needs to work normally in dark places. Therefore, a camera with infrared function is used. When the lighting conditions are insufficient, the camera will turn on the infrared fill light function to ensure that the road image can be collected normally. At the same time, the target camera is installed in the blind spot of driving vision, for example, the top of the front of the car is used to shoot the blind spot in front of the car, and the right side of the car body is used to shoot the right blind spot and other driving blind spots.

[0060] It should be noted that the position of pedestrians in road images can be located through deep learning target detection.

[0061] The preset semantic segmentation network in the vehicle terminal processes the road surface images captured by the target camera. The UNET module of the residual module of the preset semantic segmentation network is used to infer the road boundaries separated by curbs, isolation barriers, green belts and other curbs to obtain a road surface segmentation map. Different blind spot danger areas are divided according to the road surface segmentation map.

[0062] The blind spot area where the pedestrian is located can be determined based on the pedestrian's position, and pedestrians in different blind spot danger areas can be warned by traversing the area.

[0063] The blind spot pedestrian detection method provided by the present invention can accurately and real-time detect the dangerous areas in the driving vision blind spot, and give an alarm prompt to pedestrians when they enter the blind spot dangerous area, which can reduce the occurrence of safety accidents.

[0064] Further, in one embodiment, step S1 may specifically include:

[0065] S11. Detect the road surface image based on the object detection yolov5 algorithm to obtain the position of pedestrians.

[0066] Optionally, after the in-vehicle terminal obtains the road surface image collected by the target camera, it will send the road surface image to the target detection network and the preset semantic segmentation network respectively. Among them, the target detection network is implemented by the yolov5 algorithm with faster inference time, and the object detection yolov5 algorithm is used to detect the road surface image to obtain the position of pedestrians in the road surface image.

[0067] The blind spot pedestrian detection method provided by the present invention can quickly and accurately locate pedestrians in the road surface image based on the yolov5 algorithm, and then can give an alarm prompt to pedestrians in a timely and accurate manner when they enter the blind spot dangerous area, thereby reducing the occurrence of safety accidents.

[0068] Further, in one embodiment, step S2 may specifically include:

[0069] S21. Perform linear fitting on the road surface segmentation map to obtain the road edge fitting line;

[0070] S22. Determine different blind spots according to the road edge fitting line and the alarm lines preset at different positions of the vehicle body calibrated in the target camera.

[0071] Further, in one embodiment, step S21 may specifically include:

[0072] S211. Denoise the road surface segmentation map based on erosion and dilation processing;

[0073] S212. Cluster the target pixel points in the denoised road surface segmentation map according to the preset spatial clustering algorithm;

[0074] S213. Determine the target cluster containing the most target pixel points according to the clustering result;

[0075] S214. If the ratio of the number of target pixel points in the target cluster to the total number of pixel points in the road surface segmentation map is greater than or equal to the preset value, perform linear fitting on the target pixel points in the target cluster to obtain the road edge fitting line.

[0076] Optionally, after the vehicle-mounted terminal obtains the road surface image collected by the target camera, it will send the road surface image into a preset semantic segmentation network, process the road surface image collected by the target camera, use the unet module of the residual module of the preset semantic segmentation network to infer the road boundary separated by road edges such as curbs, isolation fences, and green belts, and obtain a road surface segmentation map.

[0077] Based on the least squares method, perform linear fitting on the road surface segmentation map to obtain a fitted road edge fitting line, and then determine different blind areas according to the obtained road edge fitting line and the alarm lines calibrated at different positions of the vehicle body in the preset target camera.

[0078] It should be noted that the calibration of the alarm lines at different positions of the vehicle body can be specifically carried out in the following ways:

[0079] 1. The terminal, such as a mobile phone, accesses a wireless network card, such as a USB wireless network card, to turn on the wireless connection. After the mobile phone is connected, it enters the calibration interface through the application program, selects the camera channel to be calibrated, and the terminal transmits the position of the alarm line of the corresponding channel to the mobile phone. The user can calibrate the alarm line by dragging the two ends of the line.

[0080] 2. Select the calibration mode on the multimedia intelligent central control screen, and calibrate the alarm line by dragging the two ends of the line on the corresponding camera channel screen.

[0081] If the alarm lines at different positions of the vehicle body are not pre-calibrated, the default configuration is adopted.

[0082] The process of performing linear fitting on the obtained road surface segmentation map is as follows:

[0083] Step 1: The road surface segmentation map inferred by the vehicle-mounted terminal is reflected in the form of black and white pixel points. The pixel value 0 represents the background, and 255 represents the road edge. Perform an opening operation of erosion first and then dilation on the road surface segmentation map obtained by the vehicle-mounted terminal to remove some noise in the road surface segmentation map.

[0084] Step 2: According to the preset spatial clustering DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm, cluster the target pixel points in the denoised road surface segmentation map. For example, for all points with pixel values not equal to 0 in the obtained denoised road surface segmentation map, use the DBSCAN algorithm for clustering. Approximate the road edge as a straight line. During driving, the change in the slope of this straight line in the picture is smooth. Some reference can be made to the slope of the previous fitting straight line for the current fitting process to remove some noise, and based on Step 3, determine whether any two points with pixel values not equal to 0 can be clustered into the same cluster.

[0085] Step 3: Assume that (ax, ay) and (bx, by) are the coordinates of any two points (point a and point b) among the target pixel points to be clustered respectively, and k′ is the slope of the curb fitting line obtained after the previous linear fitting. Let the slope k of the curb fitting line obtained after the current linear fitting of the road surface segmentation map be expressed as:

[0086] k = (ay - ax) / (by - bx)

[0087] If the product of k′ and k is not -1, that is, points a and b are not perpendicular to the curb fitting line obtained from the previous linear fitting, then the included angle c between the curb fitting lines obtained from the two linear fittings can be expressed as:

[0088] c = arctan(|(k - k′) / (1 + k·k′)|)

[0089] Finally, the included angle c is used as a weight to participate in the calculation of the distance between points a and b, and the distance d between points a and b is obtained:

[0090]

[0091] When the distance d is less than the preset threshold, points a and b can be clustered into the same cluster.

[0092] Step 4: According to the clustering result, select the target cluster with the largest number of target pixel points after clustering. When the proportion of the number of target pixel points in the target cluster is less than a preset value of the total number of pixel points included in the road surface segmentation map, such as 1%, it is regarded as noise and the fitting fails; when the proportion of the number of target pixel points in the target cluster is not less than a preset value of the total number of pixel points included in the road surface segmentation map, such as 1%, then linear fitting is performed on the target pixel points inside the target cluster to obtain the point - slope form of the straight line, and the final curb fitting line is obtained according to the point - slope form of the straight line.

[0093] Step 5: In order to remove the curb fitting line obtained from accidental jumps, perform Kalman filtering on each parameter involved in the point - slope form of the straight line to filter the straight lines with accidental jumps. When the fitting fails continuously for 5 times, reset the Kalman filter and reset the slope of the curb fitting line obtained from the previous effective linear fitting (that is, when calculating the distance between any two target pixel points, remove the influence of the slope).

[0094] The blind - area pedestrian detection method provided by the present invention divides different blind - area dangerous areas by using the curb fitting line and the pre - calibrated warning line, laying a foundation for subsequent real - time and accurate implementation of different alarm prompts for pedestrians in different blind areas.

[0095] Further, in one embodiment, step S3 may specifically include:

[0096] S31. If the position of the pedestrian is within the range of the area adjacent to the vehicle body and the first warning line, determine the blind area where the pedestrian is located as the first blind area, and give a first-level warning prompt to the pedestrian in the first blind area;

[0097] S32. If the position of the pedestrian is within the range of the area adjacent to the vehicle body and the second warning line, determine the blind area where the pedestrian is located as the second blind area, and give a second-level warning prompt to the pedestrian in the second blind area;

[0098] S33. If the position of the pedestrian is within the range of the area adjacent to the vehicle body and the third warning line, determine the blind area where the pedestrian is located as the third blind area, and give a third-level warning prompt to the pedestrian in the third blind area;

[0099] Among them, the area adjacent to the vehicle body is determined according to the area formed between the road edge fitting line and the vehicle body;

[0100] The first warning line is determined according to the warning line at the first position from the vehicle body;

[0101] The second warning line is determined according to the warning line at the second position from the vehicle body;

[0102] The third warning line is determined according to the warning line at the third position from the vehicle body.

[0103] Optionally, when the vehicle-mounted terminal is powered on, it will read the warning lines calibrated at different positions from the vehicle body in the preset target camera, and divide different warning lines according to the specific positions of the warning lines from the vehicle body. Specifically, the warning line at the first position from the vehicle body (for example, 1 meter from the vehicle body) is used as the first warning line, the warning line at the second position from the vehicle body (for example, 2 meters from the vehicle body) is used as the second warning line, and the warning line at the third position from the vehicle body (for example, 3 meters from the vehicle body) is used as the third warning line.

[0104] Judge the target box with the position of the pedestrian framed based on the target detection yolov5 algorithm, and judge whether the position of the pedestrian or one of the corners of the target box is within the range of the area adjacent to the vehicle body (the area formed between the road edge fitting line and the vehicle body) and the first warning line. If so, determine the blind area where the pedestrian is located as the first blind area, give a first-level warning prompt to the pedestrian in the first blind area, and generate a first-level alarm message.

[0105] When the position of the pedestrian or one of the corners of the target box is within the range of the above-mentioned area adjacent to the vehicle body and the second warning line, determine the blind area where the pedestrian is located as the second blind area, give a second-level warning prompt to the pedestrian in the second blind area, and generate a second-level alarm message.

[0106] When the position of the pedestrian or one corner of the target frame is within the range of the above-mentioned adjacent vehicle body area and the third alarm line, the blind spot where the pedestrian is located is determined as the third blind spot, and a third-level warning prompt is given to the pedestrian in the third blind spot, and a third-level alarm information is generated.

[0107] When the position of the pedestrian or one corner of the target frame is outside the range of the above-mentioned area adjacent to the vehicle body or outside the range of the third warning line, it is considered that the pedestrian is not in the dangerous area of the blind spot and no alarm information is generated.

[0108] Finally, the vehicle terminal sends the road image transmitted by the target camera, the obtained pedestrian position and the roadside fitting line to the Figure 2 The multimedia intelligent central control screen shown draws the position of pedestrians and roadside fitting lines on the screen of the target camera and displays them so that the driver can clearly observe the blind spot situation.

[0109] The blind spot pedestrian detection method provided by the present invention can accurately and in real time detect dangerous areas in the driving blind spot, and provide corresponding alarm prompts according to the degree of danger faced by pedestrians in different blind spots. It can reduce the occurrence of safety accidents while reducing the impact of frequent alarms on drivers.

[0110] Furthermore, in one embodiment, after step S3, the following steps may be specifically included:

[0111] S4. Send different first prompt instructions to the multimedia intelligent central control screen installed in the vehicle according to the warning level of the warning prompt:

[0112] If the alarm prompt is a level 1 alarm prompt, a first prompt instruction of repeating the first target voice a first preset number of times is sent to the multimedia intelligent central control screen;

[0113] If the alarm prompt is a level 2 alarm prompt, a first prompt instruction for repeating the first target voice a second preset number of times is sent to the multimedia intelligent central control screen;

[0114] If the alarm prompt is a level three alarm prompt, a first prompt instruction for repeating the first target voice a third preset number of times is sent to the multimedia intelligent central control screen.

[0115] Optionally, when an alarm prompt is generated, a corresponding alarm message will be generated, and the vehicle terminal will send an alarm to the Figure 2 The multimedia intelligent central control screen sends different first prompt instructions:

[0116] When the alarm prompt is a first-level alarm prompt or the alarm information is a first-level alarm message: Send a first prompt command to the multimedia intelligent central control screen to repeat the first target voice for the first preset number of times. For example, repeat the "beep" sound 4 times per second. After receiving the first prompt command of repeating the "beep" sound 4 times per second, the multimedia intelligent central control screen repeats the "beep" sound 4 times per second.

[0117] When the alarm prompt is a second-level alarm or the alarm information is a second-level alarm message: Send a first prompt command to the multimedia intelligent central control screen to repeat the first target voice for the second preset number of times. For example, repeat the "beep" sound 2 times per second. After receiving the first prompt command of the "beep" sound 2 times per second, the multimedia intelligent central control screen repeats the "beep" sound 2 times per second.

[0118] When the alarm prompt is a third-level alarm or the alarm information is a third-level alarm message: Send a first prompt command to the multimedia intelligent central control screen to repeat the first target voice for the third preset number of times. For example, repeat the "beep" sound 1 time per second. After receiving the first prompt command of the "beep" sound 1 time per second, the multimedia intelligent central control screen repeats the "beep" sound 1 time per second.

[0119] The blind area pedestrian detection method provided by the present invention can accurately and real-time detect the dangerous area in the driving vision blind area, and give alarm prompts to the driver and pedestrians when pedestrians enter the blind area dangerous area, reducing the occurrence of safety accidents.

[0120] Further, in one embodiment, after step S3, it may specifically include:

[0121] S5. According to the alarm level where the alarm prompt is located, send different second prompt commands to the sound and light alarm installed on the vehicle:

[0122] If the alarm prompt is a first-level alarm prompt, send a second prompt command to the sound and light alarm to turn on the light flashing and repeat the second target voice;

[0123] If the alarm prompt is a second-level alarm prompt, send a second prompt command to the sound and light alarm to turn on the light flashing and repeat the third target voice;

[0124] If the alarm prompt is the third-level alarm prompt, send a second prompt command to the sound and light alarm to turn on the light flashing and repeat the fourth target voice.

[0125] Optionally, when an alarm prompt is generated, corresponding alarm information will be generated. The vehicle-mounted terminal sends different second prompt commands to the Figure 2 sound and light alarm in the middle according to the alarm level corresponding to different alarm prompts:

[0126] When the alarm prompt is a first-level alarm prompt or the alarm information is first-level alarm information: send a second prompt instruction to the audible and visual alarm to turn on the flashing light and repeat the second target voice, for example, turn on the flashing light and repeat the voice prompt of "Danger". After receiving the second prompt instruction to turn on the flashing light and repeat the voice prompt of "Danger", the audible and visual alarm turns on the flashing light and repeats the voice prompt of "Danger".

[0127] When the alarm prompt is a second-level alarm or the alarm information is second-level alarm information: send a second prompt instruction to the audible and visual alarm to turn on the flashing light and repeat the third target voice, for example, turn on the flashing light and repeat the voice prompt of "Danger, keep away". After receiving the second prompt instruction to turn on the flashing light and repeat the voice prompt of "Danger, keep away", the audible and visual alarm turns on the flashing light and repeats the voice prompt of "Danger, keep away".

[0128] When the alarm prompt is a third-level alarm or the alarm information is third-level alarm information: send a second prompt instruction to the audible and visual alarm to turn on the flashing light and repeat the fourth target voice, for example, turn on the flashing light and repeat the voice prompt of "Danger from large vehicle, keep away". After receiving the second prompt instruction to turn on the flashing light and repeat the voice prompt of "Danger from large vehicle, keep away", the audible and visual alarm turns on the flashing light and repeats the voice prompt of "Danger from large vehicle, keep away".

[0129] The blind area pedestrian detection method provided by the present invention can accurately and real-time detect the dangerous areas in the driving vision blind area, and when a pedestrian enters the blind area dangerous area, the audible and visual alarm is used to give an alarm prompt to the pedestrian in the blind area dangerous area, reducing the occurrence of safety accidents.

[0130] The blind area pedestrian detection system provided by the present invention will be described below. The blind area pedestrian detection system described below can be mutually corresponding and referred to the blind area pedestrian detection method described above.

[0131] Figure 3 is a schematic structural diagram of the blind area pedestrian detection system provided by the present invention, as Figure 3 shown, including: a pedestrian detection module 310, a blind area determination module 311, and an alarm prompt module 312;

[0132] The pedestrian detection module 310 is used to detect the road surface image collected by the target camera to obtain the position of the pedestrian;

[0133] The blind area determination module 311 is used to input the road surface image into a preset semantic segmentation network to obtain a road surface segmentation map, and determine different blind areas according to the road surface segmentation map;

[0134] An alarm prompt module 312 is used to determine the blind area where a pedestrian is located according to the position of the pedestrian, and give an alarm prompt to pedestrians in different blind areas;

[0135] Among them, the target camera is installed in the blind area of the driving vision.

[0136] The blind area pedestrian detection system provided by the present invention can accurately and real-time detect the dangerous area in the blind area of the driving vision, and give an alarm prompt to pedestrians when they enter the dangerous area of the blind area, which can reduce the occurrence of safety accidents.

[0137] Figure 4 It is a schematic physical structure diagram of an electronic device provided by the present invention. As Figure 4 shown, the electronic device may include: a processor 410, a communication interface 411, a memory 412, and a bus 413. Among them, the processor 410, the communication interface 411, and the memory 412 complete mutual communication through the bus 413. The processor 410 can call the logical instructions in the memory 412 to execute the following methods:

[0138] Detect the road surface image collected by the target camera to obtain the position of the pedestrian;

[0139] Input the road surface image into a preset semantic segmentation network to obtain a road surface segmentation map, and determine different blind areas according to the road surface segmentation map;

[0140] Determine the blind area where the pedestrian is located according to the position of the pedestrian, and give an alarm prompt to pedestrians in different blind areas;

[0141] Among them, the target camera is installed in the blind area of the driving vision.

[0142] In addition, when the logical instructions in the above-mentioned memory are implemented in the form of a software functional unit and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer power supply screen (which may be a personal computer, a server, or a network power supply screen, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM, Read-only Memory), random access memory (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0143] Furthermore, the present invention discloses a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the blind area pedestrian detection method provided by each of the above method embodiments, for example, including:

[0144] Detect the road surface image collected by the target camera to obtain the position of the pedestrian;

[0145] Input the road surface image into a preset semantic segmentation network to obtain a road surface segmentation map, and determine different blind areas according to the road surface segmentation map;

[0146] Determine the blind area where the pedestrian is located according to the position of the pedestrian, and give an alarm prompt to the pedestrians in different blind areas;

[0147] Wherein, the target camera is installed in the blind area of the driving vision.

[0148] On the other hand, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is implemented to execute the blind area pedestrian detection method provided by each of the above embodiments, for example, including:

[0149] Detect the road surface image collected by the target camera to obtain the position of the pedestrian;

[0150] Input the road surface image into a preset semantic segmentation network to obtain a road surface segmentation map, and determine different blind areas according to the road surface segmentation map;

[0151] Determine the blind area where the pedestrian is located according to the position of the pedestrian, and give an alarm prompt to the pedestrians in different blind areas;

[0152] Wherein, the target camera is installed in the blind area of the driving vision.

[0153] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0154] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the above technical solution, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer power supply screen (which can be a personal computer, a server, or a network power supply screen, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0155] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A blind area pedestrian detection method, characterized in that, Including: Detecting the road surface image collected by the target camera to obtain the position of pedestrians; Inputting the road surface image into a preset semantic segmentation network to obtain a road surface segmentation map, and determining different blind areas according to the road surface segmentation map; Determining the blind areas where pedestrians are located according to the positions of the pedestrians, and giving warning prompts to pedestrians in different blind areas; Wherein, the target camera is installed in the blind area of the driving vision; The determining different blind areas according to the road surface segmentation map includes: Performing linear fitting on the road surface segmentation map to obtain a road edge fitting line; In the performing linear fitting on the road surface segmentation map to obtain a road edge fitting line, it includes: Performing denoising processing on the road surface segmentation map based on erosion and dilation processing; Clustering the target pixel points in the denoised road surface segmentation map according to a preset spatial clustering algorithm; Also including: (ax, ay) and (bx, by) are respectively the coordinates of any two points (point a and point b) among the target pixel points to be clustered, k' is the slope of the road edge fitting line obtained after the previous linear fitting. Let the slope k of the road edge fitting line obtained after the current linear fitting of the road surface segmentation map be expressed as: k = (ay - ax) / (by - bx) If the product of k' and k is not -1, that is, points a and b are not perpendicular to the road edge fitting line obtained from the previous linear fitting, then calculate the included angle c between the road edge fitting lines obtained from the two linear fittings, which is expressed as: c = arctan((k - k') / (1 + k * k')) Adding the included angle c as a weight to the calculation of the distance between points a and b, and obtaining the distance d between points a and b: When the distance d is less than a preset threshold, then cluster points a and b into the same cluster; The performing linear fitting on the road surface segmentation map to obtain a road edge fitting line also includes: Determining a target cluster containing the most of the target pixel points according to the clustering result; If the ratio of the number of target pixel points in the target cluster to the total number of pixel points in the road surface segmentation map is greater than or equal to a preset value, then perform linear fitting on the target pixel points in the target cluster to obtain the road edge fitting line.

2. The blind area pedestrian detection method according to claim 1, wherein, The detecting the road surface image collected by the target camera to obtain the position of pedestrians includes: Detecting the road surface image based on the target detection yolov5 algorithm to obtain the position of the pedestrians.

3. The blind area pedestrian detection method according to claim 1, characterized in that, The determining different blind areas according to the road surface segmentation map further includes: Determining different blind areas according to the road edge fitting line and the alarm lines calibrated at different positions of the vehicle body in the preset target camera.

4. The blind area pedestrian detection method according to claim 3, wherein, The determining the blind areas where pedestrians are located according to the positions of the pedestrians, and giving warning prompts to pedestrians in different blind areas includes: If the position of the pedestrian is within the range of the area near the vehicle body and the first alarm line, then determine the blind area where the pedestrian is located as the first blind area, and give a first-level warning prompt to the pedestrians in the first blind area; If the position of the pedestrian is within the range of the area near the vehicle body and the second alarm line, then determine the blind area where the pedestrian is located as the second blind area, and give a second-level warning prompt to the pedestrians in the second blind area; If the position of the pedestrian is within the range of the adjacent vehicle body area and the third warning line, determine that the blind area where the pedestrian is located is the third blind area, and give a three-level warning prompt to the pedestrian in the third blind area; Among them, the adjacent vehicle body area is determined according to the area formed between the road edge fitting line and the vehicle body; The first warning line is determined according to the warning line at the first position from the vehicle body; The second warning line is determined according to the warning line at the second position from the vehicle body; The third warning line is determined according to the warning line at the third position from the vehicle body.

5. The blind area pedestrian detection method according to claim 4, characterized in that, After determining the blind area where the pedestrian is located according to the position of the pedestrian and giving a warning prompt to the pedestrian in different blind areas, it further includes: According to the warning level of the warning prompt, send different first prompt commands to the multimedia intelligent central control screen installed on the vehicle: If the warning prompt is the first-level warning prompt, send a first prompt command to repeat the first target voice for the first preset number of times to the multimedia intelligent central control screen; If the warning prompt is the second-level warning prompt, send a first prompt command to repeat the first target voice for the second preset number of times to the multimedia intelligent central control screen; If the warning prompt is the third-level warning prompt, send a first prompt command to repeat the first target voice for the third preset number of times to the multimedia intelligent central control screen.

6. The blind area pedestrian detection method according to claim 4, characterized in that After determining the blind area where the pedestrian is located according to the position of the pedestrian and giving a warning prompt to the pedestrian in different blind areas, it further includes: According to the warning level of the warning prompt, send different second prompt commands to the sound and light alarm installed on the vehicle: If the warning prompt is the first-level warning prompt, send a second prompt command to turn on the light flashing and repeat the second target voice to the sound and light alarm; If the warning prompt is the second-level warning prompt, send a second prompt command to turn on the light flashing and repeat the third target voice to the sound and light alarm; If the warning prompt is the third-level warning prompt, send a second prompt command to turn on the light flashing and repeat the fourth target voice to the sound and light alarm.

7. A blind area pedestrian detection system, characterized in that, It includes: A pedestrian detection module, a blind area determination module, and a warning prompt module; The pedestrian detection module is used to detect the road surface image collected by the target camera to obtain the position of the pedestrian; The blind area determination module is used to input the road surface image into a preset semantic segmentation network to obtain a road surface segmentation map, and determine different blind areas according to the road surface segmentation map; The warning prompt module is used to determine the blind area where the pedestrian is located according to the position of the pedestrian and give a warning prompt to the pedestrian in different blind areas; Among them, the target camera is installed in the driving vision blind area; Determining different blind areas according to the road surface segmentation map includes: Performing linear fitting on the road surface segmentation map to obtain a road edge fitting line; In the process of performing linear fitting on the road surface segmentation map to obtain a road edge fitting line, it includes: Performing denoising processing on the road surface segmentation map based on erosion and dilation processing; Clustering the target pixel points in the denoised road surface segmentation map according to a preset spatial clustering algorithm; It further includes: (ax, ay) and (bx, by) are the coordinates of any two points (point a and point b) among the target pixel points to be clustered respectively, k' is the slope of the road edge fitting line obtained after the previous linear fitting. Let the slope k of the road edge fitting line obtained by linearly fitting the current road surface segmentation map be expressed as: k = (ay - ax) / (by - bx) If the product of k' and k is not -1, that is, points a and b are not perpendicular to the road edge fitting line obtained by the previous linear fitting, then calculate the included angle c between the road edge fitting lines obtained by the two linear fittings, which is expressed as: c = arctan((k - k') / (1 + k * k')) Add the included angle c as a weight to the calculation of the distance between points a and b, and obtain the distance d between points a and b: When the distance d is less than the preset threshold, cluster points a and b into the same cluster; The linear fitting of the road surface segmentation map to obtain the road edge fitting line further includes: Determine the target cluster containing the most of the target pixel points according to the clustering result; If the ratio of the number of target pixel points in the target cluster to the total number of pixel points in the road surface segmentation map is greater than or equal to the preset value, then perform linear fitting on the target pixel points in the target cluster to obtain the road edge fitting line.

8. An electronic device, comprising a processor and a memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the blind area pedestrian detection method according to any one of claims 1 to 6.

9. A processor-readable storage medium, characterized in that, The processor-readable storage medium stores a computer program, and the computer program is used to cause the processor to execute the steps of the blind area pedestrian detection method according to any one of claims 1 to 6.

Images