A target detection false detection and missing detection processing method and device
By acquiring information from multiple consecutive frames of images during object detection, the labeled area of missed targets is expanded and similarity is calculated. Combining road coordinate intervals and vehicle aspect ratios, falsely detected targets are identified and deleted. This solves the problem of missed and false detections in existing algorithms, improves detection accuracy and recall, and reduces labeling costs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- WUHAN KOTEI INFORMATICS
- Filing Date
- 2023-04-09
- Publication Date
- 2026-07-21
AI Technical Summary
Existing target detection algorithms suffer from false positives and false negatives when identifying small targets at a distance, targets occluded by median strips, targets occluded by vehicles, and targets partially truncated in the field of view, leading to increased manpower and time costs for secondary annotation.
By acquiring target detection information from multiple consecutive frames of images, the coordinate information of the target is determined, the annotation area of candidate missed targets is expanded, image similarity is calculated, missed targets are identified, and falsely detected targets are deleted based on road coordinate intervals and vehicle aspect ratio.
It improved the accuracy and recall of the target detection algorithm, saved manpower and time costs, and improved the efficiency of target detection data annotation.
Smart Images
Figure CN116630930B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of image processing and deep learning technology, and specifically relates to a method and apparatus for handling false detection and missed detection in target detection. Background Technology
[0002] With the development of artificial intelligence technology, computer vision algorithms have been widely used in various industries. Object detection technology is one such branch. Current object detection algorithms are improving in performance, with the latest algorithm being YOLOv7. However, the problem of false positives and false negatives still exists. Taking vehicle recognition on highways as an example, even the latest open-source object detection algorithms still have issues with false positives and false negatives in situations such as small distant targets, oncoming targets obscured by medians, targets that are mutually obscured by vehicles, and targets that are partially truncated in the field of view. The current object detection annotation business is booming. Whether object detection uses pre-annotation or manual annotation methods, false positives and false negatives are inevitable, resulting in the manpower and time costs of secondary annotation. If there were a way to solve the problems of false positives and false negatives in object detection algorithms or the false negatives and false negatives left over from manual annotation, it would greatly improve the accuracy and recall of object detection and save the manpower and material costs of annotation business. Summary of the Invention
[0003] To address the problem of false positives and false negatives in target detection, a first aspect of the present invention provides a method for handling false positives and false negatives in target detection, comprising: acquiring multiple consecutive frames of images after target detection and target detection information for each frame; determining the coordinate information of each target based on the target detection information of each frame; the target detection information including the labeled region and labeled information of each target; extracting one or more candidate false negative targets based on the number of targets in adjacent frames of images for each target; expanding the labeled region of each candidate false negative target in adjacent frames of images and calculating the image similarity between the labeled region of the candidate false negative target and the labeled region of adjacent frames of images; identifying the false negative targets based on the calculated image similarity; determining the coordinate interval of the road in which each target is located based on the coordinate information of each target; and identifying and deleting one or more false positive targets based on the coordinate interval of the road and a preset aspect ratio of the vehicle.
[0004] In some embodiments of the present invention, the step of extracting one or more candidate missed targets based on the number of targets in adjacent multi-frame images of each target includes: comparing two adjacent frames of images, and taking one or more missing targets in the image with fewer targets as candidate missed targets.
[0005] Furthermore, the step of expanding the labeled region of each candidate missed target in adjacent multi-frame images and calculating the image similarity between the labeled region of the candidate missed target and the labeled region of the adjacent multi-frame images includes: using the labeled region of the candidate missed target in its labeled image as a reference, expanding the reference labeled region at the corresponding position in the adjacent multi-frame images through geometric transformation; and calculating the image similarity between the labeled region of the candidate missed target and the labeled region of the adjacent multi-frame images.
[0006] In some embodiments of the present invention, determining the coordinate range of the road where each target is located based on its coordinate information includes: determining the coordinate range of the road where the vehicle is located based on the geometric relationship between the vehicle and the target vehicle, and the trajectory of the target vehicle.
[0007] Furthermore, the step of determining and deleting one or more false detection targets based on the coordinate range of the road and the preset aspect ratio of the vehicle includes: determining and deleting one or more false detection targets based on whether the coordinates of each target are within the coordinate range of the road; determining and deleting one or more false detection targets based on whether the coordinates of each target conform to the ratio of the longitudinal coordinate distance between its own vehicle and the target vehicle to the height of the target vehicle; and determining and deleting one or more false detection targets based on whether the pixel length and width of each target conform to the preset aspect ratio of the vehicle.
[0008] In the above embodiments, obtaining the target detection information of each of the consecutive multi-frame images after target detection includes: performing target detection on the consecutive multi-frame images using a target detection method to obtain the target detection information of each of the consecutive multi-frame images after target detection.
[0009] A second aspect of the present invention provides a target detection false positive and false negative processing device, comprising: an acquisition module, configured to acquire a series of consecutive frames of images after target detection and target detection information for each frame; determine the coordinate information of each target based on the target detection information of each frame; the target detection information includes a labeled region and labeled information for each target; a false negative processing module, configured to extract one or more candidate false negative targets based on the number of targets in adjacent frames of images for each target; expand the labeled region of each candidate false negative target in adjacent frames of images and calculate the image similarity between the labeled region of the candidate false negative target and the labeled region of adjacent frames of images; identify the false negative target based on the image similarity calculation result; and a false positive processing module, configured to determine the coordinate interval of the road in which each target is located based on the coordinate information of each target; and determine and delete one or more false positive targets based on the coordinate interval of the road and a preset aspect ratio of the vehicle.
[0010] A third aspect of the present invention provides an electronic device, comprising: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the target detection false detection and missed detection processing method provided in the first aspect of the present invention.
[0011] In a fourth aspect, the present invention provides a computer-readable medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the target detection false detection and missed detection processing method provided in the first aspect of the present invention.
[0012] The beneficial effects of this invention are:
[0013] This invention provides a method for handling missed and false detections in target detection. It extracts frames from dashcam videos of highways or urban roads and converts them into continuous frame images. It uses rules such as the relationship between target distance and target size and a strategy that maximizes the similarity of images of the same target to regenerate missed target boxes and delete falsely detected target boxes. This improves the accuracy and recall of the target detection algorithm, effectively enhances the practical application capability of the target detection algorithm and the efficiency of target detection data annotation, and greatly saves manpower and time costs. Attached Figure Description
[0014] Figure 1 This is a schematic diagram of the basic process of the target detection false detection and false detection handling method in some embodiments of the present invention;
[0015] Figure 2 This is a schematic diagram of the specific process of the target detection false detection and false detection handling method in some embodiments of the present invention;
[0016] Figure 3 This is a schematic diagram illustrating the principle of missed detection in target detection in some embodiments of the present invention;
[0017] Figure 4 This is a schematic diagram illustrating the principle of false positives and false negatives in target detection in some embodiments of the present invention;
[0018] Figure 5 These are images before the missed detection processing in some embodiments of the present invention;
[0019] Figure 6 These are images processed by the target detection false positive and false negative processing method in some embodiments of the present invention;
[0020] Figure 7 These are images before false detection processing in some embodiments of the present invention;
[0021] Figure 8These are images processed by the false detection handling method based on target detection false detection and false detection handling in some embodiments of the present invention;
[0022] Figure 9 This is a schematic diagram of the target detection false detection and false detection processing device in some embodiments of the present invention;
[0023] Figure 10 This is a schematic diagram of the structure of an electronic device in some embodiments of the present invention. Detailed Implementation
[0024] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.
[0025] refer to Figure 1 and Figure 2 In a first aspect of the present invention, a method for handling false positives and false negatives in target detection is provided, comprising: S100. acquiring a series of consecutive frames of images after target detection and target detection information for each frame; determining the coordinate information of each target based on the target detection information of each frame; the target detection information includes the labeled region and labeled information of each target; S200. extracting one or more candidate false negative targets based on the number of targets in the adjacent frames of images for each target; expanding the labeled region of each candidate false negative target in the adjacent frames, and calculating the image similarity between the labeled region of the candidate false negative target and the labeled region of the adjacent frames; identifying the false negative target based on the calculation result of the image similarity; S300. determining the coordinate interval of the road in which each target is located based on the coordinate information of each target; determining and deleting one or more false positive targets based on the coordinate interval of the road and the preset aspect ratio of the vehicle.
[0026] In step S100 of some embodiments of the present invention, multiple consecutive frames of images after target detection and target detection information of each frame are acquired; the coordinate information of each target is determined based on the target detection information of each frame; the target detection information includes the labeled area and labeled information of each target.
[0027] Specifically, the video from a dashcam on a highway or urban road is first extracted frame by frame and converted into continuous frame images. The latest object detection algorithm, YOLOv7, is then used to detect the images, generating 2D bounding boxes and saving the coordinates of each target vehicle in a TXT or other file format. Alternatively, manually labeled images that have not been re-checked will also have corresponding TXT or XML files storing the 2D bounding box coordinates for each image. It can be understood that the object detection information includes the labeled region and labeling information for each target. The labeled region is usually a rectangular box (label box or anchor box), or varies according to the target's geometry. Correspondingly, in a road scene, the vehicle's labeling information includes the vehicle's category attributes (person, vehicle, or obstacle, etc.) and the vehicle's coordinates in the image reference system or image coordinate system. The coordinate information can also be obtained through coordinate transformation between the odometer, navigation positioning device, and image capturing device (dashcam, vision sensor, or camera, etc.).
[0028] In step S200 of some embodiments of the present invention, extracting one or more candidate missed targets based on the number of targets in adjacent multi-frame images of each target includes: S201. Comparing two adjacent frames of images, and taking one or more missing targets in the image with fewer targets as candidate missed targets.
[0029] Furthermore, the step of expanding the labeled region of each candidate missed target in adjacent multi-frame images and calculating the image similarity between the labeled region of the candidate missed target and the labeled region of adjacent multi-frame images includes: S202. Selecting the labeled region of the missed target in its labeled image as a reference, expanding the reference labeled region at the corresponding position in adjacent multi-frame images through geometric transformation; S203. Calculating the image similarity between the labeled region of the candidate missed target and the labeled region of adjacent multi-frame images.
[0030] Specifically, for the missed targets, firstly, the txt file of each image 'a' in the image storage path is read to obtain the coordinate information of each target. Then, the txt file of the previous or next image 'b' is read to obtain the number of targets in the two image txt files respectively. When the number of targets in image 'b' is one less than the number of targets in image 'a', it is assumed that the missing target in image 'a' is numbered 1 in image 'b'. Based on the target's coordinates Pa(p1, p2, p3, p4), the img.crop method of the pil library is used to crop and save the target 1. Then, in the adjacent frame images, the coordinates (p1, p2, p3, p4) are used as the center, and the coordinates are moved horizontally, vertically, diagonally, zoomed in, and zoomed out in units of one or more pixels to obtain Pb(p1', p2', p3', p4'). The images are cropped according to these coordinates, and the similarity between the image and the previously cropped image is calculated.
[0031] Image similarity calculation methods can use both traditional and deep learning methods. Traditional methods include cosine similarity, hash algorithms, and SSIM structural similarity. Open-source deep learning methods include SimGNN, VGGNet, and SiameseNetwork. A similarity threshold of 0.5 and an IoU threshold of 0.5 are used; values below these thresholds are discarded. IoU is the ratio of the overlapping area of two bounding boxes to the area of the bounding box itself. This is because the positional coordinates of the same target in adjacent images are not significantly different, meaning the IoU value is close to 1. If the IoU is too low, the probability that the two bounding boxes are the same target is low. The coordinates Pc of the cropped image with the highest similarity are selected. If the similarity is too low, the probability that the two bounding boxes are the same target is low. Then, cv2.rectangle is used to automatically draw bounding boxes according to these coordinates and describe the target type with text. The target coordinate Pc information is saved and updated in the txt file of the original image b. This generates the missed target boxes and their coordinates in image b. After multiple iterations, all missed targets can be identified. The illustration above illustrates this. Figure 2 As shown. The principle is that the same target has the highest probability of appearing near the corresponding position in adjacent frames, and the images have the highest similarity.
[0032] Understandably, the above detection strategy, for missed targets, first reads the txt file of each image in the image storage path to obtain the coordinate information of each target. Then, it reads the txt files of the previous and next images of that image. When the number of targets decreases by one, the target is cropped and saved according to its coordinates. Then, in adjacent frames, the target is moved horizontally, vertically, diagonally, zoomed in, and zoomed out in units of one or more pixels, centered on the coordinates. The image is cropped according to the coordinates, and the similarity between the image and the previously cropped image is calculated. The coordinates of the image with the highest similarity are selected, and the target information is automatically drawn according to the coordinates and saved to the original txt file. After multiple iterations, all missed targets can be identified. The principle is that the same target has the highest probability of appearing near the corresponding position in adjacent frames and the highest image similarity.
[0033] In step S300 of some embodiments of the present invention, determining the coordinate range of the road where each target is located based on the coordinate information of each target includes: determining the coordinate range of the road where the vehicle is located based on the geometric relationship between the vehicle and the target vehicle, and the trajectory of the target vehicle.
[0034] Furthermore, the step of determining and deleting one or more false detection targets based on the coordinate range of the road and the preset aspect ratio of the vehicle includes: determining and deleting one or more false detection targets based on whether the coordinates of each target are within the coordinate range of the road; determining and deleting one or more false detection targets based on whether the coordinates of each target conform to the ratio of the longitudinal coordinate distance between its own vehicle and the target vehicle to the height of the target vehicle; and determining and deleting one or more false detection targets based on whether the pixel length and width of each target conform to the preset aspect ratio of the vehicle.
[0035] Specifically, for falsely detected targets: a. First, based on the coordinate information in the txt file, calculate the correlation between the target vehicle's longitudinal coordinates and the difference between the road endpoint coordinates and the target vehicle's size. Normally, the closer the target vehicle is to the road endpoint, the smaller it is. The ratio of the pixel distance between the target vehicle and the road endpoint to the target vehicle's width is usually around 1. b. Calculate the aspect ratio of all target vehicles. Normally, the aspect ratio of a vehicle is around 1.5. c. Read the coordinate information of all target vehicles and roughly determine the coordinate range of the road based on the target vehicle's coordinate movement trajectory. Then, read the txt files of all images and delete and update the coordinate information of targets that do not conform to the road coordinate range, do not conform to the target vehicle's aspect ratio, and do not conform to the correlation between the difference between the target vehicle's longitudinal coordinates and the road endpoint coordinates and the target vehicle's size. This removes the falsely detected targets, leaving only the target bounding boxes that meet the requirements. Then, use cv2.rectangle to automatically draw bounding boxes for the remaining target coordinates, thus deleting the falsely detected targets.
[0036] refer to Figure 4 For: 1) A target not appearing in the lane is definitely not the target vehicle; 2) Target vehicles, whether large or small, must meet a certain width-to-height ratio. If the ratio is seriously inconsistent, it is definitely not the target vehicle; 3) The farther away the target vehicle is, the smaller it appears. If the target appears abnormally large or small at the current distance, it is definitely not the target vehicle.
[0037] Understandably, for falsely detected targets, the process begins by using the coordinate information in the txt file: the ratio of the longitudinal distance between the vehicle and the target vehicle to the target vehicle's height, as well as the target vehicle's aspect ratio. Based on the target vehicle's trajectory, the approximate coordinate range of the road is determined. Then, the txt files of all images are read, and the coordinate information of targets that do not conform to the road coordinate range, the target vehicle's aspect ratio, or the ratio of the longitudinal distance between the vehicle and the target vehicle to the target vehicle's height is deleted and updated in the txt file. This process removes the falsely detected targets.
[0038] Example 2
[0039] refer to Figure 9In a second aspect, the present invention provides a target detection false positive and false negative processing device 1, comprising: an acquisition module 11, configured to acquire a series of consecutive frames of images after target detection and target detection information for each frame; determine the coordinate information of each target based on the target detection information of each frame; the target detection information includes the labeled region and labeled information of each target; a false negative processing module 12, configured to extract one or more candidate false negative targets based on the number of targets in the adjacent frames of images for each target; expand the labeled region of each candidate false negative target in the adjacent frames of images, and calculate the image similarity between the labeled region of the candidate false negative target and the labeled region of the adjacent frames of images; identify the false negative target based on the image similarity calculation result; and a false positive processing module 13, configured to determine the coordinate interval of the road in which each target is located based on the coordinate information of each target; and determine and delete one or more false positive targets based on the coordinate interval of the road and the preset aspect ratio of the vehicle.
[0040] Furthermore, the missed detection processing module 12 includes: an extraction unit, used to extract one or more candidate missed targets based on the number of targets in the adjacent multi-frame images of each target; an expansion unit, used to expand the labeled area of each candidate missed target in the adjacent multi-frame images, and calculate the image similarity between the labeled area of the candidate missed target and the labeled area of the adjacent multi-frame images; and an identification unit, used to identify the missed targets based on the calculation results.
[0041] Example 3
[0042] refer to Figure 10 A third aspect of the present invention provides an electronic device comprising: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method of the first aspect of the present invention.
[0043] Electronic device 500 may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 501, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 502 or a program loaded from storage device 508 into random access memory (RAM) 503. The RAM 503 also stores various programs and data required for the operation of electronic device 500. The processing unit 501, ROM 502, and RAM 503 are interconnected via bus 504. An input / output (I / O) interface 505 is also connected to bus 504.
[0044] Typically, the following devices can be connected to I / O interface 505: input devices 506 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 507 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 508 including, for example, hard disks; and communication devices 509. Communication device 509 allows electronic device 500 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 10 An electronic device 500 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 10 Each box shown can represent a device or multiple devices as needed.
[0045] Specifically, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a storage device 508, or installed from a ROM 502. When the computer program is executed by a processing device 501, it performs the functions defined in the methods of embodiments of this disclosure. It should be noted that the computer-readable medium described in embodiments of this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.
[0046] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more computer programs, which, when executed by the electronic device, cause the electronic device to:
[0047] Computer program code for performing the operations of embodiments of this disclosure can be written in one or more programming languages or a combination thereof. Programming languages include object-oriented programming languages—such as Java, Smalltalk, C++, and Python—and conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).
[0048] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.
[0049] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for handling false positives and false negatives in target detection, characterized in that, include: Acquire multiple consecutive frames of images after object detection and object detection information for each frame; Based on the target detection information of each frame of the image, determine the coordinate information of each target; The target detection information includes the labeled area and labeled information for each target; Based on the number of targets in the adjacent multi-frame images of each target, extract one or more candidate missed targets; expand the labeled area of each candidate missed target in the adjacent multi-frame images, and calculate the image similarity between the labeled area of the candidate missed target and the labeled area of the adjacent multi-frame images; identify the missed targets based on the image similarity calculation results. Based on the coordinate information of each target, determine the coordinate range of the road in which it is located; based on the coordinate range of the road and the preset aspect ratio of the vehicle, identify and delete one or more false detection targets: based on whether the coordinates of each target are within the coordinate range of the road, identify and delete one or more false detection targets; based on whether the coordinates of each target conform to the ratio of the longitudinal distance between its own vehicle and the target vehicle to the height of the target vehicle, identify and delete one or more false detection targets; based on whether the pixel length and width of each target conform to the preset aspect ratio of the vehicle, identify and delete one or more false detection targets. The step of determining the coordinate range of the road where each target is located based on its coordinate information includes: determining the coordinate range of the road where the vehicle is located based on the geometric relationship between the vehicle and the target vehicle, as well as the trajectory of the target vehicle.
2. The target detection false detection and false negative detection handling method according to claim 1, characterized in that, The step of extracting one or more candidate missed targets based on the number of targets in adjacent multi-frame images of each target includes: Compare two adjacent frames and select the missing target or object in the image with fewer targets as a candidate for missed detection.
3. The target detection false detection and false negative detection handling method according to claim 2, characterized in that, The step of expanding the labeled region of each candidate missed target in adjacent multi-frame images and calculating the image similarity between the labeled region of the candidate missed target and the labeled regions of adjacent multi-frame images includes: Using the labeled area of the candidate missed target in its labeled image as a reference, the reference labeled area at the corresponding position in adjacent multiple frames is expanded through geometric transformation; Calculate the image similarity between the labeled region of the candidate missed target and the labeled regions of multiple adjacent frames.
4. The target detection false detection and false negative detection processing method according to any one of claims 1 to 3, characterized in that, The acquisition of consecutive multi-frame images after target detection and the target detection information of each frame includes: The target detection method is used to perform target detection on the continuous multi-frame images to obtain the continuous multi-frame images after target detection and the target detection information of each frame image.
5. A target detection false detection and false detection processing device, characterized in that, include: The acquisition module is used to acquire multiple consecutive frames of images after target detection and the target detection information of each frame; Based on the target detection information of each frame of the image, determine the coordinate information of each target; The target detection information includes the labeled area and labeled information for each target; The missed detection processing module is used to extract one or more candidate missed targets based on the number of targets in the adjacent multi-frame images of each target; expand the labeled area of each candidate missed target in the adjacent multi-frame images, and calculate the image similarity between the labeled area of the candidate missed target and the labeled area of the adjacent multi-frame images; and identify the missed targets based on the image similarity calculation results. The false detection handling module is used to determine the coordinate range of the road where each target is located based on its coordinate information; based on the road coordinate range and the vehicle's preset aspect ratio, it identifies and deletes one or more false detection targets: it identifies and deletes one or more false detection targets based on whether the coordinates of each target are within the road coordinate range; it identifies and deletes one or more false detection targets based on whether the coordinates of each target conform to the ratio of the longitudinal distance between its own vehicle and the target vehicle to the height of the target vehicle; and it identifies and deletes one or more false detection targets based on whether the pixel length and width of each target conform to the vehicle's preset aspect ratio. The step of determining the coordinate range of the road where each target is located based on its coordinate information includes: determining the coordinate range of the road where the vehicle is located based on the geometric relationship between the vehicle and the target vehicle, as well as the trajectory of the target vehicle.
6. The target detection false detection and missed detection processing device according to claim 5, characterized in that, The missed detection processing module includes: The extraction unit is used to extract one or more candidate missed targets based on the number of targets in the adjacent multiple frames of each target; An expansion unit is used to expand the labeled region of each candidate missed target in multiple adjacent frames of images, and to calculate the image similarity between the labeled region of the candidate missed target and the labeled region of the multiple adjacent frames of images. The identification unit is used to identify missed targets based on the calculation results.
7. An electronic device, comprising: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the target detection false detection and missed detection processing method as described in any one of claims 1 to 4.
8. A computer-readable medium having a computer program stored thereon, wherein, When the computer program is executed by the processor, it implements the target detection false detection and missed detection processing method as described in any one of claims 1 to 4.