Anti-collision head alarm probability determination method and device, electronic equipment and storage medium

Abstract

The present disclosure provides a collision avoidance head alarm probability determination method and device, electronic equipment and storage medium. The method comprises: acquiring a front row image of a vehicle and a rear row image of the vehicle; identifying a target region in the front row image of the vehicle, and a matched human head not located in the front seat and the aisle, to obtain a target head bounding box; the target region refers to a risk region in which the head may collide with an obstacle; determining an original alarm probability according to the size comparison relationship between the area of the target head bounding box and the area of a preset head bounding box; identifying the human body and the head in the rear row image of the vehicle to obtain the number of human bodies in the second row seat and the number of heads matched with the human bodies in the second row seat; determining an alarm probability correction value according to the number matching relationship between the number of human bodies in the second row seat and the number of heads; and correcting the original alarm probability by using the alarm probability correction value to obtain a target alarm probability.

Description

Technical Field

[0001] This disclosure relates to the field of anti-collision head technology, and in particular to a method, device, electronic device and storage medium for determining the alarm probability of an anti-collision head. Background Technology

[0002] In the relevant head-collision avoidance technology, taking the scenario of preventing a person's head from colliding with an obstacle (entertainment screen) inside the vehicle as an example: a time-of-flight (TOF) camera is usually used to obtain the depth information between the person's head and the camera. The distance between the person's head and the camera indicated by the depth information is used to determine whether the person's head has entered a target area with a risk of head collision, and then the probability of head-collision avoidance alarm is determined based on the judgment result.

[0003] However, entertainment screens are usually located between the front and second rows of seats, and TOF cameras are usually positioned 20 centimeters behind the entertainment screen. Therefore, TOF cameras cannot capture the area within 20 centimeters behind the entertainment screen, which leads to inaccurate determination of the anti-collision alarm probability by the anti-collision head technology using TOF cameras, and easily results in missed alarms. Summary of the Invention

[0004] This disclosure provides a method, apparatus, electronic device, and storage medium for determining the probability of head collision alarm.

[0005] According to a first aspect of this disclosure, a method for determining the probability of an anti-collision head alarm is provided, comprising:

[0006] Acquire front and rear images of the vehicle; the front image refers to the image acquired by the front image acquisition device located in front of the front seat in the cabin, and the rear image refers to the image acquired by the rear image acquisition device located behind an obstacle.

[0007] Identify the head of the person in the front image of the vehicle that is located in the target area and is not located in the front seat or aisle, and obtain the bounding box of the target head; the target area refers to the risk area where the head may collide with an obstacle.

[0008] The original alarm probability is determined by comparing the area of ​​the target head's outer frame with the area of ​​the preset head's outer frame.

[0009] Identify human bodies and heads in the rear seat image of a vehicle to obtain the number of human bodies in the second row and the number of heads that match the human bodies in the second row.

[0010] The alarm probability correction value is determined based on the matching relationship between the number of people and the number of heads in the second row of seats;

[0011] The original alarm probability is corrected using the alarm probability correction value to obtain the target alarm probability.

[0012] In some embodiments of this disclosure, identifying the head of a person in a front-row image of a vehicle that is located in a target area and whose matched human body is not located in the front seat or aisle, and obtaining the bounding box of the target head, includes:

[0013] Identify human bodies and heads in the front row image of a vehicle, and obtain the bounding box of each human body and the bounding box of each head in the front row image of the vehicle.

[0014] Determine the intersection-union ratio (IUR) of each person's bounding box and each person's head bounding box. Match the bounding boxes of people and heads with IUR that are not less than a preset IUR threshold to obtain the matching results.

[0015] Based on the position coordinates of the human body external frame inside the vehicle, determine the human body external frame with position coordinates in the front seats and aisle, including;

[0016] Delete the bounding boxes of the heads that match the bounding boxes of the heads in the front seats and aisle from the bounding boxes of the matching results to obtain the bounding boxes of the target heads.

[0017] In some embodiments of this disclosure, the area of ​​the preset head outer frame includes the area of ​​the first preset head outer frame and the area of ​​the second preset head outer frame, wherein the area of ​​the second preset head outer frame is greater than the area of ​​the first preset head outer frame.

[0018] Based on the comparison between the area of ​​the target head's outer frame and the area of ​​the preset head's outer frame, the original alarm probability is determined, including:

[0019] Multiply the length and width of the bounding box of the target head to obtain the area of ​​the bounding box of the target head;

[0020] Determine whether the area of ​​the target head bounding box is smaller than the area of ​​the first preset head bounding box;

[0021] If the judgment result is negative, then

[0022] The first difference value is obtained by subtracting the area of ​​the target head bounding box from the area of ​​the first preset head bounding box.

[0023] The second difference value is obtained by subtracting the area of ​​the second preset head outline from the area of ​​the first preset head outline.

[0024] The ratio of the first difference to the second difference is taken as the first ratio.

[0025] The original alarm probability is obtained by summing the first ratio with the preset alarm probability correction coefficient.

[0026] In some embodiments of this disclosure, an alarm probability correction value is determined based on the matching relationship between the number of people and the number of heads in the second row of seats, including:

[0027] When the number of people in the second row of seats is 0, the alarm probability correction value is determined to be the first correction value;

[0028] When the number of people in the second row of seats is greater than the number of heads, the alarm probability correction value is determined to be the second correction value;

[0029] When the number of people in the second row of seats equals the number of heads, the alarm probability correction value is determined to be the third correction value; where the first and second correction values ​​are positive, and the first correction value is less than the second correction value; the third correction value is negative.

[0030] In some embodiments of this disclosure, before identifying human bodies and heads in the rear seat image of a vehicle to obtain the number of human bodies in the second row and the number of heads matching the human bodies in the second row, the method for determining the probability of the anti-collision head alarm provided in this disclosure includes:

[0031] Determine the orientation of the target head bounding box relative to the target area based on the position coordinates of the bounding box of the target head.

[0032] Identify human bodies and heads in the rear seat image of a vehicle to obtain the number of human bodies in the second row and the number of heads that match the human bodies in the second row, including:

[0033] Identify human bodies and heads in the rear seat image of a vehicle to obtain the number of human bodies in the same position in the second row of seats, and the number of heads that match the human bodies in the same position in the second row of seats.

[0034] In some embodiments of this disclosure, before identifying the head of a person in the front image of a vehicle that is located in the target area and whose matched human body is not located in the front seat or aisle, and obtaining the bounding box of the target human head, the method for determining the probability of the anti-collision head alarm provided in this disclosure includes:

[0035] Acquire multiple training images, where each training image refers to an image showing the boundary position of a human head within the risk area;

[0036] The heads in the training images are identified to obtain multiple bounding boxes for the heads in the training images; each head corresponds to one bounding box.

[0037] Determine the center point of the bounding box for each head to obtain the center point positions of multiple head bounding boxes;

[0038] The target area is defined as the closed region formed by connecting the locations of multiple center points.

[0039] In some embodiments of this disclosure, the location coordinates of the human external frame within the vehicle are determined based on the location coordinates of the human external frame in the front seat and aisle, including:

[0040] Based on the position coordinates of the human body bounding box inside the vehicle, the area selected by each human body bounding box is extracted from the front row image of the vehicle to obtain at least one human body image.

[0041] Identify key points in each human body image and obtain the coordinates of the key points in each human body image;

[0042] Based on the key point coordinates of each human body image, determine the bounding box of the human body in the front seats and aisle.

[0043] According to a second aspect of this disclosure, a device for determining the probability of a collision avoidance head alarm is provided, comprising:

[0044] The acquisition unit is used to acquire images of the front row and the rear row of the vehicle; wherein, the front row image refers to the image acquired by the front row image acquisition device set in front of the front seat in the cabin, and the rear row image refers to the image acquired by the rear row image acquisition device set behind the obstacle.

[0045] The recognition unit is used to identify the head of a person in the front row image of the vehicle that is located in the target area and is not located in the front seat or aisle, and obtain the bounding box of the target head; the target area refers to the risk area where the head may collide with an obstacle;

[0046] The first determining unit is used to determine the original alarm probability based on the comparison between the area of ​​the target head outline and the area of ​​the preset head outline.

[0047] The second determining unit is used to identify human bodies and heads in the rear row image of the vehicle, and to obtain the number of human bodies in the second row and the number of heads that match the human bodies in the second row.

[0048] The third determining unit is used to determine the alarm probability correction value based on the matching relationship between the number of people and the number of heads in the second row of seats.

[0049] The fourth determining unit is used to correct the original alarm probability using the alarm probability correction value to obtain the target alarm probability.

[0050] According to a third aspect of this disclosure, an electronic device is provided, comprising:

[0051] At least one processor; and

[0052] A memory that is communicatively connected to at least one processor; wherein,

[0053] The memory stores instructions that can be executed by at least one processor, such that the at least one processor is able to perform the method described in the first aspect above.

[0054] According to a fourth aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are used to cause a computer to perform the method of the first aspect described above.

[0055] This disclosure provides a method, apparatus, electronic device, and storage medium for determining the probability of a collision avoidance head alarm. The method includes: acquiring images of the front and rear seats of a vehicle; wherein the front image refers to an image acquired by a front-seat image acquisition device positioned in front of the front seats in the cabin, and the rear image refers to an image acquired by a rear-seat image acquisition device positioned behind an obstacle; identifying heads in the front image that are located in a target area and whose matched human bodies are not located in the front seats or aisle, obtaining a target head bounding box; the target area refers to the risk area where the head may collide with the obstacle; determining an initial alarm probability based on a comparison between the area of ​​the target head bounding box and a preset head bounding box area; identifying human bodies and heads in the rear image to obtain the number of human bodies in the second row and the number of heads matching the human bodies in the second row; determining an alarm probability correction value based on the matching relationship between the number of human bodies and the number of heads in the second row; and correcting the initial alarm probability using the alarm probability correction value to obtain a target alarm probability.

[0056] According to the scheme disclosed herein, by identifying the head of a person in the front row image of a vehicle that is located in the target area and whose matched human body is not located in the front seat or aisle, the bounding box of the target human head is obtained. This allows the acquisition of the image between the obstacle and the rear row image acquisition device, thus obtaining the original alarm probability. Furthermore, by using the matching relationship between the number of human bodies and the number of human heads in the second row of seats, an alarm probability correction value is determined. The original probability is then corrected using the correction value, which can improve the accuracy of the alarm and avoid missed alarms.

[0057] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0058] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0059] Figure 1 This is a flowchart illustrating the method for determining the probability of an anti-collision head alarm provided in an embodiment of this disclosure;

[0060] Figure 2 This is a structural schematic diagram of the cockpit seating distribution provided in an embodiment of this disclosure;

[0061] Figure 3 This is a structural schematic diagram of key points of the human body provided in an embodiment of this disclosure;

[0062] Figure 4 This is a schematic diagram of the anti-collision head alarm probability determination device provided in an embodiment of this disclosure;

[0063] Figure 5 A schematic block diagram of an example electronic device provided for embodiments of this disclosure. Detailed Implementation

[0064] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0065] The present disclosure provides a method for determining the probability of a collision avoidance alarm. This method can be applied to vehicles equipped with infrared (IR) cameras. The execution subject of the method can be a processor located on the vehicle or in the cloud, such as the processor of the collision avoidance alarm module in the vehicle, or the vehicle controller.

[0066] like Figure 1 As shown, the method for determining the alarm probability of the anti-collision head provided in this embodiment includes the following steps:

[0067] Step 101: Obtain images of the front and rear seats of the vehicle;

[0068] In one embodiment, the vehicle front image refers to an image acquired by a front image acquisition device located in front of the front seats in the cabin.

[0069] In one embodiment, the front image acquisition device is a camera capable of acquiring images of the vehicle interior, such as an IR camera. This disclosure does not limit the specific image acquisition device.

[0070] In one embodiment, the front image acquisition device is used to acquire images of the front seats, second-row seats and aisle area inside the vehicle. Therefore, taking an IR camera as an example, the front IR camera can be set in front of or behind the rearview mirror inside the vehicle.

[0071] In one embodiment, the front-row image acquisition device is a non-depth image acquisition device. Compared with a non-depth image acquisition device, when processing the depth image acquired by a depth image acquisition device, it is usually necessary to convert the depth information in the depth image into distance information between the target (human body, human head) and the depth image acquisition device in the depth image, which consumes a lot of computational resources.

[0072] In one embodiment, the seating layout within the vehicle cabin is as follows: Figure 2 As shown, the front row seats refer to seats 0 and 1, the second row seats refer to seats 2 and 3, the third row seats refer to seats 4 and 5, and the aisle seat refers to seat 7.

[0073] In one embodiment, since the front image acquisition device can be located in front of or behind the rearview mirror, or at the top of the vehicle, the camera of the front image acquisition device is at a downward angle. If the user in the front seat leans back to the area around the obstacle (entertainment screen), the head of the user in the front seat will be captured by the front image acquisition device.

[0074] In one embodiment, the rear-seat image of a vehicle refers to an image acquired by a rear-seat image acquisition device positioned behind an obstacle.

[0075] In one embodiment, the rear image acquisition device may be the same as or different from the front image acquisition device, and this disclosure does not limit this.

[0076] In one embodiment, the rear-row image acquisition device is used to acquire images of the second row of seats inside the vehicle. Therefore, the rear-row image acquisition device can be positioned 20 centimeters in front of the obstacle.

[0077] Step 102: Identify the head of the person in the front row image of the vehicle that is located in the target area and whose matching body is not located in the front seat or aisle, and obtain the bounding box of the target head.

[0078] In one embodiment, the target area refers to the risk area where a person's head may collide with an obstacle.

[0079] In one embodiment, since the collision head alarm probability determined by this disclosure is the collision head alarm probability when a user in the second row of seats leans forward, the target area is the risk area where a user in the second row of seats may collide with an obstacle when leaning forward.

[0080] In one embodiment, a human head detection and classification network can be used to identify human bodies and heads in the front row image of a vehicle, and to obtain the bounding box of each human body and the bounding box of each head in the front row image of the vehicle.

[0081] In one embodiment, the human body image of each person is cropped from the front row image of the vehicle by combining the front row image of the vehicle and the human body outline of each person mentioned above.

[0082] In one embodiment, a human keypoint detection network can be used to identify human images of each human body and obtain the coordinate information of human keypoints in each human body image.

[0083] In one embodiment, key human body points refer to points that represent key locations on the human body, such as... Figure 3As shown.

[0084] In one embodiment, the head matching each human body is obtained by combining the head bounding box of each head and the head key point information in the coordinate information of human body key points in the human body image of each human body.

[0085] In one embodiment, heads matching each human body are selected from the front row images of the vehicle and located in the target area.

[0086] In one embodiment, each human body and its seat are matched based on the human body's external frame and seat information to obtain the human head in the front seat and aisle of the vehicle.

[0087] In one embodiment, the system further filters the front images of the vehicle to find human heads that match the target area but are not located in the front seats or aisle, and obtains the bounding box of the target human head.

[0088] Step 103: Determine the original alarm probability based on the comparison between the area of ​​the target head outline and the area of ​​the preset head outline.

[0089] In one embodiment, the area of ​​the preset head outline can be obtained from historical empirical values ​​of relevant anti-collision alarm thresholds.

[0090] In one embodiment, the preset head area threshold refers to the area value of the bounding box of the head on the front row image of the vehicle when the head is at the boundary of the risk area.

[0091] In one embodiment, the preset head area threshold may also refer to the area value of the bounding box of the human head on the front image of the vehicle when the human head comes into contact with an obstacle.

[0092] In one embodiment, the area value of the bounding box of the person's head on the front image of the vehicle when the person's head is at the boundary of the risk area, and the area value of the bounding box of the person's head on the front image of the vehicle when the person's head is in contact with an obstacle, can be used simultaneously to determine the original alarm probability.

[0093] In one embodiment, the bounding box of the target head can be rectangular, and the area of ​​the bounding box of the target head refers to the product of the length and width of the bounding box of the target head.

[0094] In one embodiment, the bounding box of the target head can be elliptical, and the area of ​​the bounding box of the target head refers to the product of the semi-major axis, the semi-minor axis and pi of the bounding box of the target head.

[0095] In one embodiment, since the front row image of the vehicle is an image captured by a non-depth image acquisition device, the size of the bounding box of the human head in the front row image of the vehicle is related to the distance between the human head and the obstacle.

[0096] In one embodiment, the anti-collision head alarm is triggered when the area of ​​the target head outer frame is equal to the area of ​​the preset head outer frame.

[0097] In one embodiment, when the area of ​​the target head's outer frame is smaller than the area of ​​the preset head's outer frame, it indicates that the head is far from the front-row image acquisition device and far from obstacles, making the possibility of a head collision small. Therefore, the corresponding original alarm probability is also small.

[0098] In one embodiment, the original alarm probability can be 0 or any positive number less than 1. For example, the original alarm probability can be 0, which is not limited in this disclosure.

[0099] In one embodiment, when the original alarm probability is less than 1, no anti-collision head alarm is issued; when the original alarm probability is not less than 1, an anti-collision head alarm is issued.

[0100] In one embodiment, when the area of ​​the target head's outer frame is smaller than the area of ​​the preset head's outer frame, it indicates that the head is closer to the front-row image acquisition device and also closer to the obstacle, making a head collision more likely. Therefore, the corresponding original alarm probability is also higher, and a head collision alarm needs to be issued.

[0101] Step 104: Identify human bodies and heads in the rear seat image of the vehicle to obtain the number of human bodies in the second row of seats and the number of heads that match the human bodies in the second row of seats.

[0102] In one embodiment, the number of people in the two rows of seats can be one or more, and this disclosure does not limit this.

[0103] In one embodiment, taking the number of people in the second row of seats as an example of 2, the number of heads matching the two people in the second row of seats is obtained.

[0104] Step 105: Determine the alarm probability correction value based on the matching relationship between the number of people and the number of heads in the second row of seats;

[0105] In one embodiment, the alarm probability correction value is used to correct the original alarm probability.

[0106] In one embodiment, when the number of people and heads in the second row of seats is 0, it indicates that the people and heads in the second row of seats are leaning forward too far, causing the rear image capture device to be unable to capture them. Therefore, it is necessary to increase the alarm probability.

[0107] In one embodiment, if the number of human bodies in the second row of seats is not less than 1 and the number of heads is 0, it indicates that the human bodies and heads in the second row of seats are not completely matched, that is, at least one actual head is stretched forward too far, so the alarm probability needs to be increased.

[0108] In one embodiment, when the number of human bodies and the number of heads in the two rows of seats are equal, it means that the human bodies and heads in the two rows of seats can be completely matched, that is, the human bodies and heads are both within the acquisition range of the rear image acquisition device, so it is necessary to reduce the probability of alarm.

[0109] Step 106: Correct the original alarm probability using the alarm probability correction value to obtain the target alarm probability.

[0110] In one embodiment, the target alarm probability is obtained by summing the alarm probability correction value with the original alarm probability.

[0111] In one embodiment, the original alarm probability is corrected using the alarm probability correction values ​​in the aforementioned cases to obtain the target alarm probability.

[0112] In one embodiment, the target alarm probability can improve the accuracy of the anti-collision head alarm.

[0113] The method for determining the probability of a collision avoidance head alarm provided in this embodiment includes: acquiring images of the front and rear rows of a vehicle; wherein, the front row image refers to an image acquired by a front image acquisition device located in front of the front seats in the cabin, and the rear row image refers to an image acquired by a rear image acquisition device located behind an obstacle; identifying the head of a person in the front row image that is located in a target area and whose matched human body is not located in the front seat or aisle, and obtaining a bounding box of the target head; the target area refers to the risk area where the head may collide with the obstacle; determining the original alarm probability based on the size comparison between the area of ​​the bounding box of the target head and the area of ​​a preset bounding box of the head; identifying the human body and head in the rear row image to obtain the number of human bodies in the second row and the number of heads that match the human bodies in the second row; determining an alarm probability correction value based on the matching relationship between the number of human bodies and the number of heads in the second row; and correcting the original alarm probability using the alarm probability correction value to obtain the target alarm probability.

[0114] According to the scheme disclosed herein:

[0115] First, by identifying the head of a person in the front row of the vehicle that is located in the target area but not in the front seat or aisle, the bounding box of the target head is obtained. This allows the acquisition of the image between the obstacle and the rear image acquisition device, thus obtaining the original alarm probability. Furthermore, by using the matching relationship between the number of people and the number of heads in the second row of seats, an alarm probability correction value is determined. This correction value is then used to correct the original probability, which can improve the accuracy of the alarm and avoid missed alarms.

[0116] In one embodiment, identifying a head in the front image of a vehicle that is located in the target area and whose matched human body is not located in the front seat or aisle, and obtaining the bounding box of the target head, includes:

[0117] Identify human bodies and heads in the front row image of a vehicle, and obtain the bounding box of each human body and the bounding box of each head in the front row image of the vehicle.

[0118] In one embodiment, the head bounding box typically refers to the smallest bounding box of the head.

[0119] In one embodiment, a human head classification and detection network can be used to identify human bodies and heads in the front row image of a vehicle to obtain the bounding box of each human body and the bounding box of each head in the front row image of the vehicle.

[0120] In one embodiment, the human body bounding box and the human head bounding box are typically represented by their corresponding position coordinates. Taking a rectangle as an example, the position coordinates of the top-left and bottom-right vertices of the rectangle are usually used to represent the corresponding human body bounding box.

[0121] In one embodiment, the human head classification and detection network can also determine the confidence level of the bounding box for each human body and the bounding box for each head. The confidence level is used to indicate the degree of trustworthiness of the human body bounding box or the bounding box for the head.

[0122] In one embodiment, a confidence threshold can be preset, and human bounding boxes and head bounding boxes with confidence levels lower than the preset confidence threshold can be deleted, leaving only reliable human bounding boxes and head bounding boxes.

[0123] Determine the intersection-union ratio (IUR) of each person's bounding box and each person's head bounding box. Match the bounding boxes of people and heads with IUR that are not less than a preset IUR threshold to obtain the matching results.

[0124] In one embodiment, regardless of whether the user's posture is sitting, standing, or other (bending over, looking down, turning their head, etc.), their human body outline and head outline typically have overlapping areas.

[0125] In one embodiment, the overlapping area of ​​the human body bounding box and the human head bounding box is defined as the intersection region, and the area occupied by the human body bounding box and the human head bounding box is defined as the union region; the intersection-union ratio refers to the ratio of the intersection region to the union region.

[0126] In one embodiment, a higher intersection-over-union (IoU) ratio indicates a greater overlap between the bounding boxes of the human body and the head, meaning a higher probability that the corresponding human body and head belong to the same user. Therefore, the IoU ratio of the bounding boxes of the human body and the head at different poses of the same user can be calibrated based on a large number of implementations and used as a preset IoU threshold.

[0127] In one embodiment, the preset intersection-union ratio threshold refers to a critical value at which the human body bounding box can be considered to match the human head bounding box.

[0128] In one embodiment, if the intersection-union ratio (CUI) of the human body bounding box and the human head bounding box is not less than a preset CUI threshold, then all human body bounding boxes and human head bounding boxes with CUIs not less than the preset CUI threshold can be matched.

[0129] In one embodiment, in the matching results, the human body and head corresponding to the matching human body bounding box and head bounding box belong to the same user.

[0130] Based on the position coordinates of the human body external frame inside the vehicle, determine the position coordinates of the human body external frame in the front seat and aisle.

[0131] In one embodiment, the position coordinates of the human external frame inside the vehicle refer to the position coordinates of the center point of the human external frame inside the vehicle.

[0132] In one embodiment, if the center point of the human body external frame is in the front seat of the vehicle, then the human body external frame is determined to be in the front seat.

[0133] Delete the bounding boxes of the heads that match the bounding boxes of the heads in the front seats and aisle from the bounding boxes of the matching results to obtain the bounding boxes of the target heads.

[0134] In one embodiment, the target head bounding box refers to the head bounding box of the user belonging to the second row of seats.

[0135] In one embodiment, by deleting the bounding box of the head that matches the bounding box of the head in the front row seat and aisle from the bounding box of the matching result, the head of the person belonging to the second row seat can be obtained, and thus the bounding box of the target head can be obtained.

[0136] In one embodiment, the area of ​​the preset head outline includes the area of ​​the first preset head outline and the area of ​​the second preset head outline, wherein the area of ​​the second preset head outline is larger than the area of ​​the first preset head outline.

[0137] In one embodiment, the first preset head area threshold refers to the area value of the bounding box of the head on the front row image of the vehicle when the head is at the boundary of the risk area.

[0138] In one embodiment, the second preset head area threshold refers to the area value of the bounding box of the head on the front image of the vehicle when the head comes into contact with an obstacle.

[0139] In one embodiment, the original alarm probability is determined based on a comparison between the area of ​​the target head's outer frame and the area of ​​a preset head's outer frame, including:

[0140] Multiply the length and width of the bounding box of the target head to obtain the area of ​​the bounding box of the target head;

[0141] In one embodiment, the outer frame of the human head can be a rectangular frame or an elliptical frame.

[0142] In one embodiment, if the outer frame of the human head is a rectangle, the area of ​​the outer frame of the target human head is obtained by multiplying the length and width of the rectangle.

[0143] In one embodiment, if the outer frame of the human head is an elliptical frame, the area of ​​the outer frame of the target human head is obtained by multiplying the major semi-axis, minor semi-axis and pi of the elliptical frame.

[0144] Determine whether the area of ​​the target head bounding box is smaller than the area of ​​the first preset head bounding box;

[0145] In one embodiment, if the area of ​​the outer frame of the target head is smaller than the area of ​​the first preset outer frame of the head, the original alarm probability is determined to be the first preset alarm probability. Preferably, the first preset alarm probability can be a number between 0 and 1.

[0146] If the judgment result is negative, then

[0147] The first difference value is obtained by subtracting the area of ​​the target head bounding box from the area of ​​the first preset head bounding box.

[0148] The second difference value is obtained by subtracting the area of ​​the second preset head outline from the area of ​​the first preset head outline.

[0149] The ratio of the first difference to the second difference is taken as the first ratio.

[0150] The original alarm probability is obtained by summing the first ratio with the preset alarm probability correction coefficient.

[0151] In one embodiment, a preset alarm probability correction coefficient is used to correct the original alarm probability.

[0152] In one embodiment, the original alarm probability can be a preset coefficient multiple of the first ratio, plus the sum of the preset alarm probability correction coefficient.

[0153] In one embodiment, the sum of the preset alarm probability correction coefficient and the preset coefficient is 1, such as the preset alarm probability correction coefficient and the preset coefficient are both 0.5, or the preset alarm probability correction coefficient is 0.4 and the preset coefficient is 0.6. In this disclosure, the specific values ​​of the preset alarm probability correction coefficient and the preset coefficient are not limited.

[0154] In one embodiment, when the area of ​​the target head outer frame is smaller than the area of ​​the second preset head outer frame, the first ratio is less than 1, and the original alarm probability is also less than 1, so there is no need to trigger the anti-collision head alarm.

[0155] In one embodiment, when the area of ​​the target head outer frame is greater than the area of ​​the second preset head outer frame, the first ratio is greater than 1, and the original alarm probability is also greater than 1, then the anti-collision head alarm will be triggered.

[0156] In one embodiment, when the area of ​​the target head outer frame is greater than the area of ​​the first preset head outer frame and less than the area of ​​the second preset head outer frame, the original alarm probability is a value between 0 and 1; when the area of ​​the target head outer frame is greater than the area of ​​the second preset head outer frame, the original alarm probability is a value not less than 1. Therefore, the value of the original alarm probability can be regarded as a continuous value that increases with the area of ​​the target head outer frame.

[0157] In one embodiment, an alarm probability correction value is determined based on the matching relationship between the number of people and the number of heads in the second row of seats, including:

[0158] When the number of people in the second row of seats is 0, the alarm probability correction value is determined to be the first correction value;

[0159] When the number of people in the second row of seats is greater than the number of heads, the alarm probability correction value is determined to be the second correction value;

[0160] When the number of people in the second row of seats equals the number of heads, the alarm probability correction value is determined to be the third correction value;

[0161] In one embodiment, the first correction value and the second correction value are positive numbers, and the first correction value is less than the second correction value.

[0162] In one embodiment, the third correction value is negative.

[0163] In one embodiment, when the number of people in the second row of seats is 0, it indicates that the people in the second row of seats are leaning forward too far, causing the rear image acquisition device to be unable to capture them. Therefore, it is necessary to increase the alarm probability. For example, the first correction value can be 0.3. In this disclosure, the specific value of the first correction value is not limited.

[0164] In one embodiment, when the number of human bodies in the second row of seats is greater than the number of heads, it indicates that the human bodies and heads in the second row of seats are not completely matched, that is, at least one actual head is leaning forward too far. Therefore, it is necessary to increase the alarm probability. For example, the second correction value can be 0.5. In this disclosure, the specific value of the second correction value is not limited.

[0165] In one embodiment, when the number of human bodies in the second row of seats is equal to the number of heads, it means that the human bodies and heads in the second row of seats can be completely matched, that is, the human bodies and heads are both within the acquisition range of the rear image acquisition device. Therefore, it is necessary to reduce the alarm probability. For example, the third correction value can be -0.2. In this disclosure, the specific value of the third correction value is not limited.

[0166] In one embodiment, before identifying human bodies and heads in the rear seat image of a vehicle to obtain the number of human bodies in the second row and the number of heads matching the human bodies in the second row, the method for determining the probability of the collision avoidance head alarm includes:

[0167] Determine the orientation of the target head bounding box relative to the target area based on the position coordinates of the bounding box of the target head.

[0168] In one embodiment, the orientation of the target area includes the left side of the target area and the right side of the target area.

[0169] In one embodiment, the position coordinates of the outer frame of the target head and the position coordinates of the target area are used to determine whether the outer frame of the target head is to the left or right of the target area.

[0170] Identify human bodies and heads in the rear seat image of a vehicle to obtain the number of human bodies in the second row and the number of heads that match the human bodies in the second row, including:

[0171] Identify human bodies and heads in the rear seat image of a vehicle to obtain the number of human bodies in the same position in the second row of seats, and the number of heads that match the human bodies in the same position in the second row of seats.

[0172] In one embodiment, if the position coordinates of the outer frame of the target head are to the left of the target area, the number of people in the leftmost second row of seats in the second row of seats and the number of heads that match the people in the leftmost second row of seats in the second row of seats are obtained.

[0173] In one embodiment, before identifying the head of a person in the front image of a vehicle that is located in the target area and whose matched human body is not located in the front seat or aisle, and obtaining the bounding box of the target human head, the method for determining the probability of the collision avoidance head alarm includes:

[0174] Acquire multiple training images, where each training image refers to an image showing the boundary position of a human head within the risk area;

[0175] In one embodiment, only three or more head positions can constitute a closed area. Therefore, the number of "multiple" positions can be three, four, or more, and this disclosure does not limit this to a specific number.

[0176] In one embodiment, the boundaries of the risk area are typically defined by the product manager.

[0177] In one embodiment, the product manager can designate relevant personnel to sit in the front seat or second-row seat, etc., to perform target area calibration. Taking the second-row seat as an example, relevant personnel can sit in two second-row seats simultaneously, or only one relevant personnel can sit in one of the second-row seats, leaning forward to the boundary of the risk area, and training images are collected by the in-vehicle IR camera.

[0178] Therefore, in one embodiment, a training image may contain one head or two heads. If a training image contains two heads, the region formed by the positions of each head in the training image is determined as the target region.

[0179] In one embodiment, the boundary of the risk area refers to the location where the collision avoidance alarm is triggered.

[0180] The heads in the training images are identified to obtain multiple bounding boxes for the heads in the training images; each head corresponds to one bounding box.

[0181] In one embodiment, a human head detection and classification network is used to identify the human head in each training image to obtain the bounding box of the human head in each training image.

[0182] Determine the center point of the bounding box for each head to obtain the center point positions of multiple head bounding boxes;

[0183] In one embodiment, the center point location is used to indicate the center coordinates of the outer frame of the human head.

[0184] In one embodiment, the center point position of the head outline can be obtained by the position of the head outline.

[0185] The target area is defined as the closed region formed by connecting the locations of multiple center points.

[0186] In one embodiment, the size of the target region is determined by the size of the region enclosed by all head positions at the boundary of the risk region in all training images.

[0187] In one embodiment, determining the location coordinates of the human external frame within the vehicle, specifically the human external frame located in the front seat and aisle, includes:

[0188] Based on the position coordinates of the human body bounding box inside the vehicle, the area selected by each human body bounding box is extracted from the front row image of the vehicle to obtain at least one human body image.

[0189] In one embodiment, based on the position coordinates of each person's external bounding box within the vehicle, a cropping operation in image processing is used to crop the human image corresponding to each person's external bounding box from the front row image of the vehicle.

[0190] Identify key points in each human body image and obtain the coordinates of the key points in each human body image;

[0191] In one embodiment, a human keypoint detection network can be used to identify human keypoints in each human body image and obtain the position coordinates of the keypoints in each human body image.

[0192] In one embodiment, the human body key points include head key points, and the position coordinates of the head key points are used to indicate the coordinate information of the head key points.

[0193] Based on the key point coordinates of each human body image, determine the bounding box of the human body in the front seats and aisle.

[0194] In one embodiment, if the key point coordinates of the human body image are in the front row seat, then the human body bounding box in the front row seat is determined.

[0195] According to the scheme disclosed herein:

[0196] First, by identifying the head of a person in the front row of the vehicle that is located in the target area but not in the front seat or aisle, the bounding box of the target head is obtained. This allows the acquisition of the image between the obstacle and the rear image acquisition device, thus obtaining the original alarm probability. Furthermore, by using the matching relationship between the number of people and the number of heads in the second row of seats, an alarm probability correction value is determined. This correction value is then used to correct the original probability, which can improve the accuracy of the alarm and avoid missed alarms.

[0197] Secondly, by deleting the head location information corresponding to heads with a confidence level lower than the preset head confidence level threshold from the head location information, the possibility of misidentifying other items placed on the seat as heads can be effectively reduced, thereby reducing the occurrence of false alarms in collision avoidance alarms.

[0198] Furthermore, since the method for determining the probability of head collision alarm in this disclosure uses the human body position information and head position information in the target image to determine the probability of collision alarm, and then decides whether to trigger the collision alarm, there are certain requirements for the resolution of the image acquisition device. In this disclosure, both the front and rear image acquisition devices are non-depth image acquisition devices, which can meet the high resolution requirements of the target image acquired by the image acquisition device in this disclosure.

[0199] Corresponding to the above-described method for determining the probability of a collision avoidance head alarm, this invention also proposes a device for determining the probability of a collision avoidance head alarm. Since the device embodiments of this invention correspond to the method embodiments described above, details not disclosed in the device embodiments can be referred to in the method embodiments described above, and will not be repeated here.

[0200] Figure 4 This is a schematic diagram of the structure of a head-collision alarm probability determination device provided in an embodiment of the present disclosure, as shown below. Figure 4 As shown, the anti-collision head alarm probability determination device 400 includes:

[0201] The acquisition unit 401 is used to acquire front row images and rear row images of the vehicle; wherein, the front row images refer to the images acquired by the front row image acquisition device set in front of the front seats in the cabin, and the rear row images refer to the images acquired by the rear row image acquisition device set behind the obstacle.

[0202] The recognition unit 402 is used to recognize the head of a person in the front row image of the vehicle that is located in the target area and whose matched human body is not located in the front seat or aisle, and to obtain the bounding box of the target human head; the target area refers to the risk area where the human head may collide with an obstacle;

[0203] The first determining unit 403 is used to determine the original alarm probability based on the comparison between the area of ​​the target head outline and the area of ​​the preset head outline.

[0204] The second determining unit 404 is used to identify human bodies and heads in the rear row image of the vehicle, and to obtain the number of human bodies in the second row and the number of heads that match the human bodies in the second row.

[0205] The third determining unit 405 is used to determine the alarm probability correction value based on the matching relationship between the number of people and the number of heads in the second row of seats.

[0206] The fourth determining unit 406 is used to correct the original alarm probability using the alarm probability correction value to obtain the target alarm probability.

[0207] In one embodiment, the identification unit 402 is specifically used for:

[0208] Identify human bodies and heads in the front row image of a vehicle, and obtain the bounding box of each human body and the bounding box of each head in the front row image of the vehicle.

[0209] Determine the intersection-union ratio (IUR) of each person's bounding box and each person's head bounding box. Match the bounding boxes of people and heads with IUR that are not less than a preset IUR threshold to obtain the matching results.

[0210] Based on the position coordinates of the human body external frame inside the vehicle, determine the position coordinates of the human body external frame in the front seat and aisle.

[0211] Delete the bounding boxes of the heads that match the bounding boxes of the heads in the front seats and aisle from the bounding boxes of the matching results to obtain the bounding boxes of the target heads.

[0212] In one embodiment, the area of ​​the preset head outline includes the area of ​​the first preset head outline and the area of ​​the second preset head outline, wherein the area of ​​the second preset head outline is larger than the area of ​​the first preset head outline.

[0213] In one embodiment, the first determining unit 403 is specifically used for:

[0214] Multiply the length and width of the bounding box of the target head to obtain the area of ​​the bounding box of the target head;

[0215] Determine whether the area of ​​the target head bounding box is smaller than the area of ​​the first preset head bounding box;

[0216] If the judgment result is negative, then

[0217] The first difference value is obtained by subtracting the area of ​​the target head bounding box from the area of ​​the first preset head bounding box.

[0218] The second difference value is obtained by subtracting the area of ​​the second preset head outline from the area of ​​the first preset head outline.

[0219] The ratio of the first difference to the second difference is taken as the first ratio.

[0220] The original alarm probability is obtained by summing the first ratio with the preset alarm probability correction coefficient.

[0221] In one embodiment, the third determining unit 405 is specifically used for:

[0222] When the number of people in the second row of seats is 0, the alarm probability correction value is determined to be the first correction value;

[0223] When the number of people in the second row of seats is greater than the number of heads, the alarm probability correction value is determined to be the second correction value;

[0224] When the number of people in the second row of seats equals the number of heads, the alarm probability correction value is determined to be the third correction value; where the first and second correction values ​​are positive, and the first correction value is less than the second correction value; the third correction value is negative.

[0225] In one embodiment, the anti-collision head alarm probability determination device 400 further includes a target area orientation determination unit, which is used for:

[0226] Determine the orientation of the target head bounding box relative to the target area based on the position coordinates of the bounding box of the target head.

[0227] In one embodiment, the second determining unit 404 is specifically used for:

[0228] Identify human bodies and heads in the rear seat image of a vehicle to obtain the number of human bodies in the same position in the second row of seats, and the number of heads that match the human bodies in the same position in the second row of seats.

[0229] In one embodiment, the anti-collision head alarm probability determination device 400 further includes a target area determination unit, which is used for:

[0230] Acquire multiple training images, where each training image refers to an image showing the boundary position of a human head within the risk area;

[0231] The heads in the training images are identified to obtain multiple bounding boxes for the heads in the training images; each head corresponds to one bounding box.

[0232] Determine the center point of the bounding box for each head to obtain the center point positions of multiple head bounding boxes;

[0233] The target area is defined as the closed region formed by connecting the locations of multiple center points.

[0234] In one embodiment, the identification unit 402 is specifically used for:

[0235] Based on the position coordinates of the human body bounding box inside the vehicle, the area selected by each human body bounding box is extracted from the front row image of the vehicle to obtain at least one human body image.

[0236] Identify key points in each human body image and obtain the coordinates of the key points in each human body image;

[0237] Based on the key point coordinates of each human body image, determine the bounding box of the human body in the front seats and aisle.

[0238] It should be noted that the foregoing explanation of the method embodiments also applies to the apparatus of this embodiment, and the principle is the same, so it is not limited in this embodiment.

[0239] According to embodiments of the present disclosure, the present disclosure also provides an electronic device and a non-transitory computer-readable storage medium storing computer instructions.

[0240] Specifically, this disclosure provides an electronic device, including:

[0241] At least one processor; and

[0242] A memory that is communicatively connected to at least one processor; wherein,

[0243] The memory stores instructions that can be executed by at least one processor, which enables the at least one processor to perform the steps of the aforementioned method for determining the probability of a collision head alarm.

[0244] This disclosure provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to perform the steps of the aforementioned method for determining the probability of a collision head alarm.

[0245] Figure 5A schematic block diagram of an example electronic device 500 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, in-vehicle devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0246] like Figure 5 As shown, the electronic device 500 includes a computing unit 501, which can perform various appropriate actions and processes based on a computer program stored in ROM (Read-Only Memory) 502 or a computer program loaded from storage unit 508 into RAM (Random Access Memory) 503. The RAM 503 can also store various programs and data required for the operation of the device 500. The computing unit 501, ROM 502, and RAM 503 are interconnected via a bus 504. An I / O (Input / Output) interface 505 is also connected to the bus 504.

[0247] Multiple components in device 500 are connected to I / O interface 505, including: input unit 506, such as keyboard, mouse, etc.; output unit 507, such as various types of monitors, speakers, etc.; storage unit 508, such as disk, optical disk, etc.; and communication unit 509, such as network card, modem, wireless transceiver, etc. Communication unit 509 allows device 500 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0248] The computing unit 501 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, CPUs (Central Processing Units), GPUs (Graphics Processing Units), various special-purpose AI (Artificial Intelligence) computing chips, various computing units running machine learning model algorithms, DSPs (Digital Signal Processors), and any suitable processor, controller, microcontroller, etc. The computing unit 501 performs the various methods and processes described above, such as the method for determining the probability of a head-collision alarm. For example, in some embodiments, the method for determining the probability of a head-collision alarm can be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as storage unit 508. In some embodiments, part or all of the computer program can be loaded and / or installed on device 500 via ROM 502 and / or communication unit 509. When the computer program is loaded into RAM 503 and executed by the computing unit 501, one or more steps of the methods described above can be performed. Alternatively, in other embodiments, the computing unit 501 may be configured to perform the aforementioned head-collision alarm probability determination method by any other suitable means (e.g., by means of firmware).

[0249] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), ASSPs (Application-Specific Standard Products), SOCs (System-on-Chips), CPLDs (Complex Programmable Logic Devices), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0250] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0251] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, RAM, ROM, EPROM (Electrically Programmable Read-Only Memory) or flash memory, optical fiber, CD-ROM (Compact Disc Read-Only Memory), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0252] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0253] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for determining the probability of an anti-collision head alarm, characterized in that, include: Acquire front and rear images of the vehicle; wherein, the front image refers to the image acquired by a front image acquisition device located in front of the front seat in the cabin, and the rear image refers to the image acquired by a rear image acquisition device located behind an obstacle. Identify the head of the person in the front image of the vehicle that is located in the target area and is not located in the front seat or aisle, and obtain the bounding box of the target head; the target area refers to the risk area where the head may collide with an obstacle; The original alarm probability is determined by comparing the area of ​​the target head's outer frame with the area of ​​the preset head's outer frame. Identify human bodies and heads in the rear seat image of the vehicle to obtain the number of human bodies in the second row of seats and the number of heads that match the human bodies in the second row of seats. Based on the matching relationship between the number of people and the number of heads in the second row of seats, the alarm probability correction value is determined; The original alarm probability is corrected using the alarm probability correction value to obtain the target alarm probability.

2. The method according to claim 1, characterized in that, The step of identifying the head of a person in the front image of the vehicle that is located in the target area and whose matched human body is not located in the front seat or aisle, and obtaining the bounding box of the target head, includes: Identify human bodies and heads in the front row image of the vehicle to obtain the bounding box of each human body and the bounding box of each head in the front row image of the vehicle. Determine the intersection-union ratio (CURRY) of each human body bounding box and each human head bounding box, and match the human body bounding boxes and human head bounding boxes whose CURRY is not less than a preset CURRY threshold to obtain a matching result; Based on the position coordinates of the human external frame within the vehicle, determine the human external frame with the position coordinates in the front seat and aisle. Delete the bounding box of the head that matches the bounding box of the head in the front seat and aisle from the bounding box of the matching result to obtain the bounding box of the target head.

3. The method according to claim 1, characterized in that, The area of ​​the preset head outer frame includes the area of ​​the first preset head outer frame and the area of ​​the second preset head outer frame, wherein the area of ​​the second preset head outer frame is greater than the area of ​​the first preset head outer frame. The step of determining the original alarm probability based on the comparison between the area of ​​the target head's outer frame and the area of ​​a preset head's outer frame includes: Multiply the length and width of the outer frame of the target head to obtain the area of ​​the outer frame of the target head; Determine whether the area of ​​the target head outline is smaller than the area of ​​the first preset head outline; If the judgment result is negative, then The first difference value is obtained by subtracting the area of ​​the target head outer frame from the area of ​​the first preset head outer frame. The difference between the area of ​​the second preset head outline and the area of ​​the first preset head outline is used to obtain the second difference value; The ratio of the first difference to the second difference is taken as the first ratio. The original alarm probability is obtained by summing the first ratio with the preset alarm probability correction coefficient.

4. The method according to claim 1, characterized in that, The step of determining the alarm probability correction value based on the matching relationship between the number of people and the number of heads in the second row of seats includes: When the number of people in the second row of seats is 0, the alarm probability correction value is determined to be the first correction value; When the number of people in the second row of seats is greater than the number of heads, the alarm probability correction value is determined to be the second correction value; When the number of human bodies in the second row of seats equals the number of heads, the alarm probability correction value is determined to be a third correction value; wherein the first correction value and the second correction value are positive numbers, and the first correction value is less than the second correction value; the third correction value is a negative number.

5. The method according to claim 1, characterized in that, Before identifying human bodies and heads in the rear seat image of the vehicle to obtain the number of human bodies in the second row of seats and the number of heads matching the human bodies in the second row of seats, the method includes: Based on the position coordinates of the outer frame of the target head, determine the orientation of the outer frame of the target head relative to the target area; The step of identifying human bodies and heads in the rear seat image of the vehicle to obtain the number of human bodies in the second row of seats and the number of heads that match the human bodies in the second row of seats includes: Identify human bodies and heads in the rear seat image of the vehicle to obtain the number of human bodies in the second row of seats at the same location, and the number of heads that match the human bodies in the second row of seats at the same location.

6. The method according to claim 1, characterized in that, Before identifying the head of a person in the front image of the vehicle that is located in the target area and whose matched human body is not located in the front seat or aisle, and obtaining the bounding box of the target head, the method includes: Acquire multiple training images, wherein the training images refer to images of a human head at the boundary position of the risk area; The heads in the training images are identified to obtain multiple bounding boxes for the heads in the training images; wherein, each head corresponds to one bounding box. The center point position of each of the human head bounding frames is determined to obtain the center point positions of multiple human head bounding frames; The closed area formed by connecting the positions of the multiple center points is defined as the target area.

7. The method according to claim 2, characterized in that, The step of determining the position coordinates of the human external frame in the front seat and aisle based on the position coordinates of the human external frame in the vehicle includes: Based on the position coordinates of the human body external frame inside the vehicle, at least one human body image is obtained by cropping the area selected by each human body external frame from the front row image of the vehicle. Identify key human points in each human image to obtain the coordinates of key points in each human image; Based on the key point coordinates of each human body image, the bounding box of the human body in the front seat and aisle is determined.

8. A device for determining the probability of an anti-collision head alarm, characterized in that, include: An acquisition unit is used to acquire images of the front row and the rear row of a vehicle; wherein, the front row image refers to an image acquired by a front row image acquisition device located in front of the front seat in the cabin, and the rear row image refers to an image acquired by a rear row image acquisition device located behind an obstacle. The recognition unit is used to identify the head of a person in the front image of the vehicle that is located in the target area and whose matched human body is not located in the front seat or aisle, and to obtain the bounding box of the target head; the target area refers to the risk area where the head may collide with an obstacle; The first determining unit is used to determine the original alarm probability based on the comparison between the area of ​​the target head outer frame and the area of ​​the preset head outer frame. The second determining unit is used to identify human bodies and heads in the rear row image of the vehicle, and to obtain the number of human bodies in the second row of seats and the number of heads that match the human bodies in the second row of seats. The third determining unit is used to determine the alarm probability correction value based on the matching relationship between the number of people and the number of heads in the two rows of seats. The fourth determining unit is used to correct the original alarm probability using the alarm probability correction value to obtain the target alarm probability.

9. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method according to any one of claims 1 to 7.

Images