A vehicle monitoring and alarming method, device, electronic equipment and vehicle
By using multiple cameras to collaboratively process images, the location of monitored objects around the vehicle is determined and their distance from the vehicle is calculated. This solves the problem of false alarms and missed alarms caused by fisheye camera distortion, and improves the accuracy and safety of vehicle monitoring.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHEJIANG GEELY HLDG GRP CO LTD
- Filing Date
- 2024-06-07
- Publication Date
- 2026-05-29
Smart Images
Figure CN118711314B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of vehicle technology, and more specifically, to a vehicle monitoring and alarm method, device, electronic equipment, and vehicle. Background Technology
[0002] Vehicle Sentry Mode is used to help car owners monitor their parked vehicles. It typically uses multiple cameras and sensors installed on the vehicle to monitor activities around the vehicle and provide relevant notifications. For example, it uses images captured by each camera to detect human distance. When the distance between a person and the vehicle is less than a preset distance threshold and exceeds a certain time, an alarm is triggered.
[0003] Currently, surround-view cameras with fisheye lenses are commonly used for image capture. However, because the images captured by fisheye cameras have significant distortion at the edges, the measurement error of human distance at the image edges is large, which can easily cause false alarms or missed alarms, affecting the normal operation of the vehicle's sentry mode. Summary of the Invention
[0004] The problem solved by this invention is how to avoid missed or false detections when human body detection is performed in vehicle sentry mode.
[0005] To address the aforementioned problems, this invention provides a vehicle monitoring and alarm method, device, electronic equipment, and vehicle.
[0006] In a first aspect, the present invention provides a vehicle monitoring and alarm method, comprising:
[0007] Multiple images of the vehicle's surroundings captured simultaneously by multiple vehicle-mounted cameras are obtained, wherein each image includes a view of the public coverage area around the vehicle;
[0008] The multiple images are processed to determine the monitoring objects located in the common coverage area in each of the images;
[0009] Based on the monitored object in each of the images and the position of the monitored object in each of the images, determine the target position of the same monitored object in each of the images;
[0010] The distance between the monitored object and the vehicle is determined based on the target location, and an alarm is issued when the distance meets a preset condition.
[0011] Optionally, processing the multiple images to determine the monitoring objects located within the common coverage area in each of the images includes:
[0012] Target detection is performed on each of the images to determine the detection results;
[0013] Based on the detection results, the monitoring objects located in the common coverage area of each of the images are determined.
[0014] Optionally, performing target detection on each of the multiple images to determine the detection result includes:
[0015] Target bounding box detection is performed on the image to determine the position of the target bounding box corresponding to the monitored object in the image;
[0016] Key point detection is performed on the image to determine the location of the key points corresponding to the monitored object in the image.
[0017] Optionally, determining the monitoring object located in the common coverage area in each of the images based on the detection result includes:
[0018] The center point position of the target bounding box is determined based on its position in the image.
[0019] The pixel value of the center point of the target bounding box at the corresponding position in the public coverage area mask is determined based on the predetermined public coverage area mask and the center point position. The pixel value is then used to determine whether the monitored object is located in the public coverage area.
[0020] In the public coverage area mask, the pixel values of pixels located in the public coverage area are different from the pixel values of pixels not located in the public coverage area.
[0021] Optionally, determining the target location of the same monitored object in each of the images based on the monitored object in each of the images and the position of the monitored object in each of the images includes:
[0022] The monitoring object image of the common coverage area in each of the images is separated by the human body detection box;
[0023] Visual feature vectors corresponding to each of the monitored object images are extracted respectively. When the similarity between the visual feature vectors corresponding to multiple monitored object images is greater than a preset similarity, the monitored object in the multiple monitored object images is determined to be the same monitored object, so as to determine the target position of the same monitored object in each of the images.
[0024] Optionally, determining the target location of the same monitored object in each of the images based on the monitored object in each of the images and the position of the monitored object in each of the images includes:
[0025] The target location of the monitored object in the image is determined based on the location of the key points in the image;
[0026] Based on a pre-determined mapping mask and the target location, determine the corresponding mapped target location in other images for the target location of the monitored object in any of the images;
[0027] When the distance between the mapped target location and the target location of the monitored object in the corresponding image meets the preset distance, the monitored object in the corresponding image is determined to be the same monitored object, so as to determine the target location of the same monitored object in each image;
[0028] The pixel values at each location in the mapping mask include the mapping relationship between the position of a location point in the image captured by one of the vehicle-mounted cameras and its position in the image captured by the other vehicle-mounted camera when the two vehicle-mounted cameras capture images around the vehicle.
[0029] Optionally, determining the distance between the same monitored object and the vehicle based on the target location includes:
[0030] The first distance between the same monitored object and the vehicle is determined based on the target position of the same monitored object in each of the images and the calibration parameters of each of the vehicle-mounted cameras;
[0031] A second distance between the same monitored object and the vehicle is determined based on the target location of the same monitored object and the distance mask of each of the vehicle-mounted cameras, wherein the pixel value of each position in the distance mask includes the distance value between the corresponding position point and the vehicle when the vehicle-mounted camera captures an image of the area around the vehicle;
[0032] The distance between the same monitored object and the vehicle is determined based on the first distance and the second distance.
[0033] In a second aspect, the present invention provides a vehicle monitoring and alarm device, comprising:
[0034] The acquisition module is used to acquire multiple images of the vehicle's surroundings captured by multiple vehicle-mounted cameras at the same time, wherein each of the images includes a view of the public coverage area around the vehicle;
[0035] A processing module is used to process multiple images to determine the monitoring objects located in the common coverage area in each of the images;
[0036] The positioning module is used to determine the target position of the same monitoring object in each of the images based on the monitoring object in each of the images and the position of the monitoring object in each of the images;
[0037] The early warning module is used to determine the distance between the same monitored object and the vehicle based on the target location, and to issue an alarm when the distance meets a preset condition.
[0038] Thirdly, the present invention provides an electronic device, including a memory and a processor;
[0039] The memory is used to store computer programs;
[0040] The processor is used to implement the above vehicle monitoring and alarm method when executing the computer program.
[0041] Fourthly, the present invention provides a vehicle including the above-mentioned electronic equipment.
[0042] This invention processes multiple images captured by multiple vehicle-mounted cameras to identify monitoring objects located within the common coverage area of the images. Then, based on the monitoring objects in each image and their positions within each image, it determines the target position of the same monitoring object in each image. This same monitoring object is used as the subject of distance detection. The distance between the same monitoring object and the vehicle is determined based on the target position. When the distance meets preset conditions, an alarm is issued, improving vehicle safety. By using multi-camera detection, the problem of "large distortion at the edges of images captured by fisheye cameras, leading to large errors in human distance measurement and easily causing false alarms or missed alarms" can be avoided, thus improving the accuracy of human detection and reducing the probability of false alarms or missed alarms. Attached Figure Description
[0043] Figure 1 This is a schematic flowchart of the vehicle monitoring and alarm method according to an embodiment of the present invention;
[0044] Figure 2 This is a schematic diagram of a camera image according to an embodiment of the present invention;
[0045] Figure 3 This is a system block diagram of the vehicle monitoring and alarm device according to an embodiment of the present invention;
[0046] Figure 4 This is a system block diagram of an electronic device according to an embodiment of the present invention;
[0047] Figure 5 This is a system block diagram of the storage medium according to an embodiment of the present invention. Detailed Implementation
[0048] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
[0049] like Figure 1 As shown, an embodiment of the present invention provides a vehicle monitoring and alarm method, including:
[0050] (1) Acquire multiple images of the vehicle’s surroundings captured by multiple vehicle cameras at the same time, wherein each of the images includes a view of the public coverage area around the vehicle.
[0051] Specifically, in combination Figure 2 As shown, taking two vehicle-mounted cameras (camera 1 and camera 2, such as the left camera and the front camera) as an example, images (including public coverage areas) can be captured at the same time, and each image includes the public coverage area around the vehicle.
[0052] Among them, the same ground point refers to the same ground point in the images of the two vehicle-mounted cameras.
[0053] (2) Process the multiple images to determine the monitoring objects in the common coverage area in each of the images.
[0054] Specifically, by processing multiple images captured by multiple vehicle-mounted cameras, such as human bounding box detection and human key point detection, the monitoring objects located in the common coverage area of multiple images can be identified.
[0055] (3) Determine the target position of the same monitoring object in each of the images based on the monitoring object in each of the images and the position of the monitoring object in each of the images.
[0056] Specifically, based on the monitored objects in each image and their positions within each image, the target position of the same monitored object in each image is determined, and this same monitored object is used as the subject of distance detection. Multi-camera detection avoids the problem that "images captured by fisheye cameras have significant distortion at the edges, leading to large errors in human distance measurement and easily causing false alarms or missed alarms," thereby improving the accuracy of human detection and reducing the probability of false alarms or missed alarms.
[0057] (4) Determine the distance between the same monitored object and the vehicle based on the target location, and issue an alarm when the distance meets the preset conditions.
[0058] Specifically, the distance between the same monitored object and the vehicle is determined based on the target location. An alarm is issued when the distance meets preset conditions, such as when the distance between the same monitored object and the vehicle is less than a preset distance threshold and remains so for more than a preset time, which can improve vehicle safety.
[0059] Optionally, processing the multiple images to determine the monitoring objects located within the common coverage area in each of the images includes:
[0060] Target detection is performed on each of the images to determine the detection results.
[0061] Specifically, target detection is performed on multiple images (e.g., images from camera 1 and camera 2), such as human bounding box detection and human key point detection, to determine the detection results for each image.
[0062] Based on the detection results, the monitoring objects located in the common coverage area of each of the images are determined.
[0063] Specifically, the detection results of each image are filtered to identify the monitored objects in the common coverage area among multiple images (e.g., images from camera 1 and camera 2), thus identifying the same monitored object. For example, a ReID (Re-Identification) method based on visual feature vectors or a location-based pairing method can be used. If either method determines that the monitored object is the same, then the monitored objects in the common coverage area detected by the two cameras are the same monitored object. This will be discussed in detail later and will not be repeated here.
[0064] Optionally, performing target detection on each of the multiple images to determine the detection result includes:
[0065] Target bounding box detection is performed on the image to determine the location of the target bounding box corresponding to the monitored object in the image.
[0066] Specifically, target bounding box detection is first performed on multiple images to determine the location (e.g., coordinates) of the target bounding box corresponding to the monitored object in the image. Target bounding box detection (e.g., the RTMDET algorithm) is used to locate and mark the position of the monitored object (e.g., a human body) in an image or video, and the specific process may include the following steps:
[0067] (1) Image preprocessing: First, the input image may need to be preprocessed to better adapt to the detection algorithm, such as adjusting the image size, color space conversion, image enhancement, etc., to improve the accuracy and robustness of the detection.
[0068] (2) Selecting a detection algorithm: Choose a human bounding box detection algorithm that suits the task requirements. Common algorithms include deep learning-based methods (such as detectors based on convolutional neural networks) and traditional machine learning methods.
[0069] (3) Feature extraction: For deep learning-based methods, features need to be extracted from the image first using a pre-trained model or a custom model. These features are usually learned through models such as convolutional neural networks, which can capture semantic information in the image.
[0070] (4) Bounding box generation: Based on feature extraction, candidate human bounding boxes are generated using certain methods (such as sliding window, region proposal network, etc.); these bounding boxes may contain some falsely detected or missed targets.
[0071] (5) Bounding box scoring: The generated bounding boxes are scored to determine which ones actually contain the human body; scoring is usually achieved by the target probability or confidence level predicted by the detection algorithm.
[0072] (6) Non-maximum suppression: For bounding boxes that may overlap, non-maximum suppression is used to filter them to eliminate redundant detection results; this process ensures that each target has only one optimal bounding box corresponding to it.
[0073] (7) Output results: Finally, the selected human bounding boxes and their corresponding confidence or probability are output; these bounding boxes are usually represented as rectangular boxes and can be used to mark the human body location in the image.
[0074] (8) Post-processing: In some cases, it may be necessary to post-process the detection results to further improve the accuracy of the detection; for example, morphological operations can be used to fill the holes in the bounding box or remove small noise areas.
[0075] Key point detection is performed on the image to determine the location of the key points corresponding to the monitored object in the image.
[0076] Specifically, keypoint detection is then performed on multiple images to determine the location (e.g., coordinates) of keypoints corresponding to the monitored object (e.g., foot keypoints) in the images. Keypoint detection is used to identify specific keypoints of the monitored object (e.g., the human body) in images or videos, such as the location of parts like the head, shoulders, arms, and knees. The specific process may include the following steps:
[0077] (1) Data preparation: First, a dataset containing human key point annotations is needed; these annotations can be manually annotated or generated by automated methods. The dataset usually includes images or videos and the corresponding key point coordinates.
[0078] (2) Selecting a detection algorithm: Choose a human keypoint detection algorithm that suits the task requirements. Common algorithms include deep learning-based methods (such as detectors based on convolutional neural networks) and traditional machine learning methods.
[0079] (3) Feature extraction: For deep learning-based methods, features need to be extracted from images using pre-trained models or custom models.
[0080] (4) Key point localization: Based on feature extraction, certain methods (such as regression, dense sampling, etc.) are used to locate the position of human key points; usually involves training a model that can predict the coordinates of each key point in a given image.
[0081] (5) Post-processing of results: For some keypoint detection algorithms, post-processing may be required to further improve the accuracy of detection. For example, smoothing techniques can be used to eliminate discontinuities or outliers in the keypoint locations.
[0082] (6) Output results: The final output is the detected human body key points and their corresponding locations; key points are usually represented as points or specific markers, which can be used to mark human body parts in images or videos.
[0083] Optionally, determining the monitoring object located in the common coverage area in each of the images based on the detection result includes:
[0084] The center point of the target bounding box is determined based on its position in the image.
[0085] Specifically, the coordinates of the center point of the target bounding box are determined based on its position in the image, for example, by using the average of the coordinates of the four corner points of the bounding box.
[0086] The pixel value of the center point of the target bounding box at the corresponding position in the public coverage area mask is determined based on the predetermined public coverage area mask and the center point position. The pixel value is then used to determine whether the monitored object is located in the public coverage area.
[0087] In the public coverage area mask, the pixel values of pixels located in the public coverage area are different from the pixel values of pixels not located in the public coverage area.
[0088] Specifically, by pre-calibrating and obtaining the mask of the common coverage area corresponding to multiple images (e.g., images from camera 1 and camera 2), and combining the coordinates of the center point, the pixel value of the center point of the target bounding box at the corresponding position in the mask of the common coverage area can be determined. If the value is 1, the monitored object is considered to be located in the common coverage area; otherwise, the monitored object is considered to be not located in the common coverage area.
[0089] The mask has the same resolution as the original image. The pixel value at each location represents whether the location is a common coverage area, with 1 representing a common coverage area and 0 representing a non-common coverage area.
[0090] Optionally, determining the target location of the same monitored object in each of the images based on the monitored object in each of the images and the position of the monitored object in each of the images includes:
[0091] The monitoring object image of the common coverage area in each of the images is separated by the human body detection box.
[0092] Specifically, the human detection box is used to separate the monitored object image from the common coverage area in multiple images (e.g., images from camera 1 and camera 2).
[0093] Visual feature vectors corresponding to each of the monitored object images are extracted respectively. When the similarity between the visual feature vectors corresponding to multiple monitored object images is greater than a preset similarity, the monitored object in the multiple monitored object images is determined to be the same monitored object, so as to determine the target position of the same monitored object in each of the images.
[0094] Specifically, the ReID model extracts visual feature vectors (such as 256-dimensional vectors) from the images of the monitored objects, and calculates the similarity of the visual feature vectors of the monitored objects in the common coverage area of multiple images (such as images from camera 1 and camera 2). When the similarity is greater than the preset similarity, the corresponding monitored objects are determined to be the same monitored object, thereby determining the target location of the same monitored object in each image.
[0095] Optionally, determining the target location of the same monitored object in each of the images based on the monitored object in each of the images and the position of the monitored object in each of the images includes:
[0096] The target location of the monitored object in the image is determined based on the position of the key points in the image.
[0097] Specifically, the target position of the monitored object in the image is obtained through key points (such as foot key points) (i.e., the average of the coordinates of two foot key points).
[0098] Based on a pre-determined mapping mask and the target location, determine the corresponding mapped target location in other images for the target location of the monitored object in any of the images.
[0099] Specifically, by pre-calibrating and obtaining a mapping mask of the coordinate positions of multiple images (e.g., images from camera 1 and camera 2) in a common area, the target positions (i.e., the human standing positions) of all target positions (e.g., human standing positions) in the common coverage area of image 1 are obtained in image 2.
[0100] When the distance between the mapped target location and the target location of the monitored object in the corresponding image meets the preset distance, the monitored object in the corresponding image is determined to be the same monitored object, so as to determine the target location of the same monitored object in each image;
[0101] The pixel values at each location in the mapping mask include the mapping relationship between the position of a location point in the image captured by one of the vehicle-mounted cameras and its position in the image captured by the other vehicle-mounted camera when the two vehicle-mounted cameras capture images around the vehicle.
[0102] Specifically, the distance between the mapped target location and all target locations in the common coverage area of the camera 2 image is calculated. When the distance meets the preset distance (e.g., the distance is less than the preset distance threshold), the corresponding monitoring object is determined to be the same monitoring object, thereby enabling the determination of the target location of the same monitoring object in each image.
[0103] The resolution of the mapping mask is the same as that of the original image. Taking two cameras as an example, the pixel value of each position includes the mapping relationship between the position of a location point in the image captured by one of the two vehicle cameras and its position in the image captured by the other vehicle camera when the two vehicle cameras capture images around the vehicle. For example, when capturing a flat ground image (preset ground image), the coordinates of the image position of the ground point captured by the ground point in the image of camera 1 and the image position of the same ground point captured by camera 2.
[0104] When there are three or more cameras, taking three cameras as an example, the same monitoring object can be identified first based on the images from camera 1 and camera 2, then another same monitoring object can be identified based on the images from camera 2 and camera 3, and finally the final same monitoring object can be found from the two identified same monitoring objects.
[0105] Optionally, determining the distance between the same monitored object and the vehicle based on the target location includes:
[0106] The first distance between the same monitored object and the vehicle is determined based on the target position of the same monitored object in each of the images and the calibration parameters of each of the vehicle-mounted cameras.
[0107] Specifically, taking two cameras as an example, the target positions of the monitored objects captured by camera 1 and camera 2 satisfy the following relationship:
[0108] (u1,v1,1)=(1 / z)*MI1*MO1*(x,y,z);
[0109] (u2,v2,1)=(1 / z)*MI2*MO2*(x,y,z);
[0110] Where u1 and v1 are the coordinates of the target position in the image of camera 1, u2 and v2 are the coordinates of the target position in the image of camera 2, x, y, and z are the three-dimensional coordinates of the target position in world coordinates, MI1 and MO1 are the intrinsic and extrinsic transformation matrices of camera 1 (obtained through calibration, i.e., calibration parameters), and MI2 and MO2 are the intrinsic and extrinsic transformation matrices of camera 2 (obtained through calibration, i.e., calibration parameters), and * represents matrix multiplication. By solving the two equations simultaneously and approximating the solution using numerical methods, the world coordinates (x, y, z) of the target position can be obtained, thus enabling the calculation of the first distance d1 from the target position to the vehicle.
[0111] When there are three or more cameras, the above relationship includes three or more equations. Taking three cameras as an example, we can first determine one solution for the target position based on the first and second equations, then determine another solution for the target position based on the second and third equations, and finally determine the world coordinates of the final target position based on the average of the two solutions.
[0112] A second distance between the same monitored object and the vehicle is determined based on the target location of the same monitored object and the distance mask of each of the vehicle-mounted cameras, wherein the pixel value of each position in the distance mask includes the distance value between the corresponding position point and the vehicle when the vehicle-mounted camera captures an image of the area around the vehicle.
[0113] Specifically, the average distance d2 between the target position of the same monitored object and the vehicle is obtained by using the distance mask of camera 1 and the distance mask of camera 2, respectively, which are pre-calibrated.
[0114] The distance mask has the same resolution as the original image. Each pixel value includes the distance between the corresponding point and the vehicle when the vehicle camera captures images of the area around the vehicle. For example, when capturing an image of a flat ground, the distance between this ground point and the vehicle is the distance.
[0115] The distance between the same monitored object and the vehicle is determined based on the first distance and the second distance.
[0116] Specifically, the average value d of the first distance d1 and the second distance d2 is calculated as the distance from the same monitored object to the vehicle.
[0117] like Figure 3 As shown, another embodiment of the present invention provides a vehicle monitoring and alarm device, comprising:
[0118] The acquisition module is used to acquire multiple images of the vehicle's surroundings captured by multiple vehicle-mounted cameras at the same time, wherein each of the images includes a view of the public coverage area around the vehicle;
[0119] A processing module is used to process multiple images to determine the monitoring objects located in the common coverage area in each of the images;
[0120] The positioning module is used to determine the target position of the same monitoring object in each of the images based on the monitoring object in each of the images and the position of the monitoring object in each of the images;
[0121] The early warning module is used to determine the distance between the same monitored object and the vehicle based on the target location, and to issue an alarm when the distance meets a preset condition.
[0122] like Figure 4 As shown, another embodiment of the present invention provides an electronic device, including a memory 402 and a processor 401 (which can be interconnected via a bus 400);
[0123] The memory is used to store computer programs;
[0124] The processor is used to implement the above vehicle monitoring and alarm method when executing the computer program.
[0125] Another embodiment of the present invention provides a vehicle including the above-described electronic equipment.
[0126] like Figure 5 As shown, another embodiment of the present invention provides a computer storage medium storing a computer program, which, when executed by a processor, implements the vehicle monitoring and alarm method described above.
[0127] While the present invention has been disclosed above, its scope of protection is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, and all such changes and modifications will fall within the scope of protection of the present invention.
Claims
1. A vehicle monitoring and alarm method, characterized in that, include: Multiple images of the vehicle's surroundings captured simultaneously by multiple vehicle-mounted cameras are obtained, wherein each image includes a view of the public coverage area around the vehicle; The multiple images are processed to determine the monitoring objects located in the common coverage area in each of the images; Based on the monitored object in each of the images and the position of the monitored object in each of the images, determine the target position of the same monitored object in each of the images; Determine the distance between the same monitored object and the vehicle based on the target location, and issue an alarm when the distance meets a preset condition; The process of processing multiple images to determine the monitoring objects located in the common coverage area in each image includes: Target detection is performed on each of the images to determine the detection results; Based on the detection results, the monitoring objects located in the common coverage area in each of the images are determined; The step of determining the monitoring object located in the common coverage area in each of the images based on the detection results includes: The center point position of the target bounding box is determined based on the position of the target bounding box corresponding to the monitored object in the image; The pixel value of the center point of the target bounding box at the corresponding position in the public coverage area mask is determined based on the predetermined public coverage area mask and the center point position. The pixel value is then used to determine whether the monitored object is located in the public coverage area. In the public coverage area mask, the pixel values of pixels located in the public coverage area are different from the pixel values of pixels not located in the public coverage area.
2. The vehicle monitoring and alarm method according to claim 1, characterized in that, The step of performing target detection on each of the multiple images to determine the detection results includes: Target bounding box detection is performed on the image to determine the position of the target bounding box corresponding to the monitored object in the image; Key point detection is performed on the image to determine the location of the key points corresponding to the monitored object in the image.
3. The vehicle monitoring and alarm method according to claim 2, characterized in that, The step of determining the target location of the same monitored object in each of the images based on the monitored object in each of the images and the position of the monitored object in each of the images includes: The monitoring object image of the common coverage area in each of the images is separated by the human body detection box; Visual feature vectors corresponding to each of the monitored object images are extracted respectively. When the similarity between the visual feature vectors corresponding to multiple monitored object images is greater than a preset similarity, the monitored object in the multiple monitored object images is determined to be the same monitored object, so as to determine the target position of the same monitored object in each of the images.
4. The vehicle monitoring and alarm method according to claim 2, characterized in that, The step of determining the target location of the same monitored object in each of the images based on the monitored object in each of the images and the position of the monitored object in each of the images includes: The target location of the monitored object in the image is determined based on the location of the key points in the image; Based on a pre-determined mapping mask and the target location, determine the corresponding mapped target location in other images for the target location of the monitored object in any of the images; When the distance between the mapped target location and the target location of the monitored object in the corresponding image meets the preset distance, the monitored object in the corresponding image is determined to be the same monitored object, so as to determine the target location of the same monitored object in each image; The pixel values at each location in the mapping mask include the mapping relationship between the position of a location point in the image captured by one of the vehicle-mounted cameras and its position in the image captured by the other vehicle-mounted camera when the two vehicle-mounted cameras capture images around the vehicle.
5. The vehicle monitoring and alarm method according to claim 1, characterized in that, Determining the distance between the same monitored object and the vehicle based on the target location includes: The first distance between the same monitored object and the vehicle is determined based on the target position of the same monitored object in each of the images and the calibration parameters of each of the vehicle-mounted cameras; A second distance between the same monitored object and the vehicle is determined based on the target location of the same monitored object and the distance mask of each of the vehicle-mounted cameras, wherein the pixel value of each position in the distance mask includes the distance value between the corresponding position point and the vehicle when the vehicle-mounted camera captures an image of the area around the vehicle; The distance between the same monitored object and the vehicle is determined based on the first distance and the second distance.
6. A vehicle monitoring and alarm device, characterized in that, include: The acquisition module is used to acquire multiple images of the vehicle's surroundings captured by multiple vehicle-mounted cameras at the same time, wherein each of the images includes a view of the public coverage area around the vehicle; A processing module is used to process multiple images to determine the monitoring objects located in the common coverage area in each of the images; The positioning module is used to determine the target position of the same monitoring object in each of the images based on the monitoring object in each of the images and the position of the monitoring object in each of the images; The early warning module is used to determine the distance between the same monitored object and the vehicle based on the target location, and to issue an alarm when the distance meets a preset condition; The process of processing multiple images to determine the monitoring objects located in the common coverage area in each image includes: Target detection is performed on each of the images to determine the detection results; Based on the detection results, the monitoring objects located in the common coverage area in each of the images are determined; The step of determining the monitoring object located in the common coverage area in each of the images based on the detection results includes: The center point position of the target bounding box is determined based on the position of the target bounding box corresponding to the monitored object in the image; The pixel value of the center point of the target bounding box at the corresponding position in the public coverage area mask is determined based on the predetermined public coverage area mask and the center point position. The pixel value is then used to determine whether the monitored object is located in the public coverage area. In the public coverage area mask, the pixel values of pixels located in the public coverage area are different from the pixel values of pixels not located in the public coverage area.
7. An electronic device, characterized in that, Including memory and processor; The memory is used to store computer programs; The processor is configured to implement the vehicle monitoring and alarm method as described in any one of claims 1-5 when executing the computer program.
8. A vehicle, characterized in that, Includes the electronic device as described in claim 7.