False detection processing method and device, electronic equipment and storage medium
The parallel architecture of AI detection and RuleBase detection performs false detection in lidar perception, solving the negative sample false detection problem caused by incomplete data sets, and achieving more accurate vehicle environment perception and decision planning.
Patent Information
- Application Number
- CN202411285092.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-13
- Publication Date
- 2024-12-13
AI Technical Summary
Lidar perception based on deep learning is limited in the data set, and negative sample misdetection is prone to negative sample detection, affecting the effect of vehicle environment perception and downstream decision planning.
Using a parallel architecture of AI detection and RuleBase detection, front and back frame matching is performed in the tracker, the characteristics of the RuleBase tracker matched on the AI tracker in the feature matching result are calculated, filter labels are added according to preset conditions, and all trackers are traversed. Delete AI detection targets that match successfully but have filtering tags.
Effectively avoid misdeleting targets, reduce false detection rates, improve the accuracy of vehicle environment perception, reduce the brakes of special vehicles in autonomous driving scenarios, and improve the ability to drive on the edge.
Smart Images
Figure CN119131561A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of laser point cloud data technology, and in particular to a false detection processing method, device, electronic device, and storage medium. Background Art
[0002] Due to the incomplete data set, the types of detectable targets for deep learning-based lidar perception are limited, and negative samples such as green plants may be mistakenly detected as targets, which will affect the results of vehicle environment perception and further affect the effectiveness of downstream decision-making and planning. Summary of the invention
[0003] The embodiments of the present application provide a false detection processing method, device, electronic device, and storage medium to handle false detection.
[0004] The present application embodiment adopts the following technical solutions:
[0005] In a first aspect, an embodiment of the present application provides a method for handling false positive detections, wherein the method comprises:
[0006] Obtaining a target for AI detection and a target for RuleBase detection, wherein the AI detection includes an artificial intelligence detection method, and the RuleBase detection includes a rule-based detection method;
[0007] According to the target detected by the AI and the target detected by the RuleBase, matching is performed in the tracker between the previous and next frames and a tracker matching result is obtained, wherein the tracker matching result includes the AI tracker and the RuleBase tracker corresponding to the real target;
[0008] Calculating the features of the RuleBase tracker matched with the AI tracker in the tracker matching result, and adding a filter tag to the AI detection target whose features meet the preset conditions; and
[0009] Traverse all trackers, find the AI detected target corresponding to the successful match with RuleBase detection, and delete the AI detected target with the filtering label.
[0010] In some embodiments, the calculating the features of the RuleBase tracker matched with the AI tracker in the tracker matching result includes:
[0011] Calculate the height value of the RuleBase detection target based on the center point and size of the RuleBase detection target;
[0012] Calculate the angle value of the RuleBase detection target according to the direction of the RuleBase detection target, where the angle value is the angle of the vector rotated counterclockwise from the positive x-axis direction of the vehicle;
[0013] According to the size of the RuleBase detection target, calculate the aspect ratio of the RuleBase detection target.
[0014] In some embodiments, adding a filter tag to an AI-detected target whose features meet preset conditions includes:
[0015] If the height value of the target detected by the RuleBase is greater than the preset threshold value, the angle value is within the angle setting range, and the aspect ratio is greater than the set threshold, the filtering label is added to the AI detected target corresponding to the target detected by the RuleBase, and the preset threshold value, the angle setting range and the set threshold are prior knowledge.
[0016] In some embodiments, the obtaining of the AI detection target and the RuleBase detection target includes:
[0017] Get the target detected by AI, including the size, center point, direction and type of the target;
[0018] Get the target detected by RuleBase, including the size, center point, direction, and type of the target.
[0019] In some embodiments, performing frame matching in a tracker based on the target detected by the AI and the target detected by the RuleBase and obtaining a tracker matching result includes:
[0020] Put the target detected by the AI and the target detected by the RuleBase into a tracking and matching module for processing, and the tracking and matching module is used to track and match the previous and next frames;
[0021] In the tracking and matching module, the target input in the first frame is used as a new target and a corresponding target tracker is generated;
[0022] Match each frame target with the target tracker and create a new tracker for unmatched targets.
[0023] In some embodiments, performing frame matching in a tracker based on the target detected by the AI and the target detected by the RuleBase and obtaining a tracker matching result includes:
[0024] Determine whether the 3D rectangular box of the target detected by AI intersects with the contour box of the target detected by RuleBase;
[0025] If there is an intersection, the trackers are matched with each other and the successfully matched tracker is obtained;
[0026] Synchronize the target’s feature information to the successfully matched tracker.
[0027] In some embodiments, all trackers are traversed to find the detected target when the target detected by RuleBase is successfully matched, and the AI detected target with the filter tag is deleted, including:
[0028] Traverse all trackers to find the corresponding AI detection target when AI detection and RuleBase detection match successfully, and delete the AI detection target that matches RuleBase and has a filter tag.
[0029] In a second aspect, an embodiment of the present application further provides a false detection processing device, wherein the device comprises:
[0030] An acquisition module, used to acquire a target for AI detection and a target for RuleBase detection, wherein the AI detection includes an artificial intelligence detection method, and the RuleBase detection includes a rule-based detection method;
[0031] A tracker matching module, used to match the previous and next frames in the tracker according to the target detected by the AI and the target detected by the RuleBase, and obtain a tracker matching result, wherein the tracker matching result includes the AI tracker and the RuleBase tracker corresponding to the real target;
[0032] a feature calculation module, used to calculate the features of the RuleBase tracker matched with the AI tracker in the tracker matching result, and add a filter tag to the AI detection target whose features meet the preset conditions; and
[0033] The traversal and elimination module is used to traverse all trackers, find the AI detected targets that match the targets detected by RuleBase successfully, and delete the AI detected targets with filter tags.
[0034] In a third aspect, an embodiment of the present application further provides an electronic device, comprising: a processor; and a memory arranged to store computer executable instructions, wherein the executable instructions, when executed, cause the processor to perform the above method.
[0035] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, which stores one or more programs. When the one or more programs are executed by an electronic device including multiple application programs, the electronic device executes the above method.
[0036] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects: obtaining the target detected by AI and the target detected by RuleBase, wherein the AI detection includes an artificial intelligence detection method, and the RuleBase detection includes a rule-based detection method. Target matching is performed by using AI detection and RuleBase detection to avoid accidental deletion of targets. According to the target detected by AI and the target detected by RuleBase, front and back frames are matched in the tracker and a tracker matching result is obtained. The tracker matching result includes the AI tracker and RuleBase tracker corresponding to the real target, and the obtained AI tracker and RuleBase tracker can be matched. The features of the RuleBase tracker matched with the AI tracker in the tracker matching result are calculated, and a filter label is added to the AI detected target whose features meet the preset conditions. Finally, by traversing all trackers, the corresponding AI detected target when the AI detection and RuleBase detection match successfully is found, and the target is eliminated according to the filter label. Through the above method, it is possible to better achieve false detection discrimination and reduce the false detection rate. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:
[0038] Figure 1 Schematic diagram of the process of the false detection processing method in the embodiment of the present application;
[0039] Figure 2 This is a schematic diagram of the structure of the false detection processing device in the embodiment of the present application;
[0040] Figure 3 It is a schematic diagram of the structure of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION
[0041] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.
[0042] The main scenarios for false detection include roadside green plants, new roadside guardrails, temporary construction barriers, untrimmed flower beds, etc., which are mistakenly detected as pedestrians, bicycles, vehicles, etc.
[0043] For more specific application scenarios, the inventors found that in some automatic driving scenarios, special vehicles such as road sweepers and minibuses are very sensitive to green plants at the road boundary when they are close to the road boundary. Once the green plants are mistakenly detected as traffic participants (pedestrians, bicycles, vehicles, etc.), it is easy to cause the vehicle to brake, stop, or even dodge around obstacles, which has a great impact on the vehicle's body feeling, ability to drive close to the edge, success rate of entering and exiting the station, success rate of pulling over, etc. In addition, false detections are prone to occur, including false detections caused by guardrails that are close to each other on both sides of the road, false detections caused by temporary construction barriers, false detections caused by flower beds that invade the lanes, etc.
[0044] The technical solutions provided by various embodiments of the present application are described in detail below in conjunction with the accompanying drawings.
[0045] The present application embodiment provides a method for handling false positives, such as Figure 1 As shown, a schematic flow chart of a false detection processing method in an embodiment of the present application is provided, and the method at least includes the following steps S110 to S140:
[0046] Step S110, obtaining the target of AI detection and the target of RuleBase detection, wherein the AI detection includes an artificial intelligence detection method, and the RuleBase detection includes a rule-based detection method.
[0047] AI detection is an artificial intelligence detection method that can detect and identify targets and their types. RuleBase is a rule-based detection method that can be used as a fallback solution when AI detection may have incomplete data sets and cannot accurately identify targets. The target type detected by RuleBase is unknown.
[0048] The parallel architecture of AI-detected targets and RuleBase-detected targets can avoid accidental deletion of targets or scenarios where AI detection cannot identify them.
[0049] Step S120, according to the target detected by the AI and the target detected by the RuleBase, the previous and next frames are matched in the tracker and a tracker matching result is obtained, wherein the tracker matching result includes the AI tracker and the RuleBase tracker corresponding to the real target.
[0050] According to the pre-input of the target detected by AI and the target detected by RuleBase, the previous and next frames are tracked and matched to obtain the tracker matching result. The tracker matching result includes the AI tracker and RuleBase tracker corresponding to the real target, and the feature information related to the detected target is also synchronously put into the tracker to obtain the RuleBase target that matches the AI detection.
[0051] Step S130, calculating the features of the RuleBase tracker that matches the AI tracker in the tracker matching results, and adding a filter tag to the AI detected target whose features meet the preset conditions.
[0052] Through the constructed tracker, the targets can be classified and matched with each other to obtain the AI tracker and RuleBase tracker corresponding to the real target.
[0053] By further calculating the features of the RuleBase tracker that matches the AI tracker in the tracker matching results, a filtering label is added to the AI target corresponding to RuleBase.
[0054] Step S140, traverse all trackers, find the AI detected target corresponding to the target detected by RuleBase when it successfully matches, and delete the AI detected target with the filter tag.
[0055] Considering the possible false detection, all trackers are traversed to find the corresponding AI detection target when AI detection and RuleBase detection match successfully, and then the target is removed according to the filtering label to complete the false detection processing. The vehicle environment perception after false detection processing is input into the downstream decision planning.
[0056] The above method can be used in scenarios where roadside green plants, new roadside guardrails, temporary construction barriers, untrimmed flower beds, etc. are mistakenly detected as pedestrians, bicycles, vehicles, etc.
[0057] Through the above method, the occurrence of false detection is reduced, so that the special vehicle can reduce the occurrence of braking during operation and improve the ability to drive close to the edge.
[0058] Through the above method, we can design multi-features to locate most of the boundary road facilities based on prior knowledge, and better identify the possible false detection of green plants, guardrails, baffles, etc.
[0059] In one embodiment of the present application, the calculating of the features of the RuleBase tracker matched with the AI tracker in the tracker matching result includes: calculating the height value of the RuleBase detection target according to the center point and size of the RuleBase detection target; calculating the angle value of the RuleBase detection target according to the direction of the RuleBase detection target, wherein the angle value is the angle of the vector rotated counterclockwise from the positive x-axis direction of the vehicle; and calculating the aspect ratio of the RuleBase detection target according to the size of the RuleBase detection target.
[0060] The false detection caused by “green plants” is taken as an example for detailed explanation.
[0061] Based on the actual vehicle test bag analysis, it is found that the geometric features of boundary green plants are relatively obvious, including that the target height is relatively small, the target direction is basically parallel to the x-axis direction of the vehicle, and the target is "long strip", that is, the length is several times the width. Based on the above artificial features, a green plant false detection discrimination scheme is provided to filter out such false detections. Those skilled in the art can understand that in addition to "green plants", it can also be applied to other boundary road facilities such as guardrails, construction barriers, flower beds, etc.
[0062] In specific implementation, the height is calculated using the center point and size of the RuleBase target:
[0063] Corresponding pseudo code: obj_height = obj->center(2)-obj->size(2).
[0064] The heading angle is calculated using the direction of the RuleBase target. The heading angle is the angle from the positive x-axis counterclockwise to the vector, i.e. the vertical angle.
[0065] Corresponding pseudocode: std::atan2(obj->direction(1),obj->direction(0)).
[0066] The aspect ratio is calculated by using the RuleBase target size matched by AI detection.
[0067] Corresponding pseudocode: obj->size(0) / obj->size(1).
[0068] Those skilled in the art will appreciate that the above pseudo code is only an example and is not intended to limit the protection scope of the present application.
[0069] In one embodiment of the present application, a filter tag is added to an AI-detected target whose features meet preset conditions, including: if the height value of the RuleBase-detected target is greater than a preset threshold value, the angle value is within an angle setting range, and the aspect ratio is greater than a set threshold value, then the filter tag is added to the AI-detected target corresponding to the RuleBase-detected target, and the preset threshold value, the angle setting range, and the set threshold value are prior knowledge.
[0070] When determining an AI-detected target whose features meet the preset conditions, multiple features are calculated for the RuleBase target that matches the AI detection. For RuleBase-detected targets whose height is greater than the set threshold, whose heading is within the set range, and whose aspect ratio is greater than the set threshold, the corresponding AI detection is marked with a filter tag.
[0071] In one embodiment of the present application, the obtaining of the target detected by AI and the target detected by RuleBase includes: obtaining the target obtained by AI detection, including the size, center point, direction and type of the target; obtaining the target obtained by RuleBase detection, including the size, center point, direction and type of the target.
[0072] In the embodiment of the present application, a false positive discrimination module is constructed, and the front input includes the target detected by AI and the target detected by RuleBase. Specifically, the target features include but are not limited to size, center point, direction, type, etc.
[0073] In one embodiment of the present application, the target detected by the AI and the target detected by the RuleBase are matched in the tracker to obtain the tracker matching result, including: putting the target detected by the AI and the target detected by the RuleBase into a tracking and matching module for processing, and the tracking and matching module is used to track and match the previous and next frames; in the tracking and matching module, the target input in the first frame is used as a new target and a corresponding target tracker is generated; each frame target is matched with the target tracker and a new tracker is created for the unmatched target.
[0074] For the tracking and matching process of the previous and next frames, all current targets will enter the tracking and matching module for processing. In the tracking and matching module, the targets input in the first frame will generate corresponding trackers as new targets, and then each frame target will come in to match the target tracker. For example, Hungarian matching can be used to put each target into a matching tracker, and then a new tracker is created for the unmatched targets. It can be understood that the above Hungarian matching is only an example and is not used to limit the scope of protection of this application.
[0075] In one embodiment of the present application, the target detected by the AI and the target detected by the RuleBase are matched in the tracker to obtain the tracker matching result, including: determining whether the 3D rectangular box of the target detected by the AI intersects with the target contour box detected by the RuleBase; if there is an intersection, matching the trackers with each other to obtain a successfully matched tracker; synchronizing the feature information of the target to the successfully matched tracker.
[0076] Based on the tracker matching, the matching relationship between the trackers is determined by judging whether the 3D rectangular box of the AI detection target and the outline box of RuleBase intersect (for example, by comparing the IOU value). After matching the trackers with each other, the AI and RuleBase trackers corresponding to the real target are obtained. At this time, the feature information such as the type of the target is synchronized with each successfully matched tracker and the feature information is saved in the corresponding tracker.
[0077] In one embodiment of the present application, all trackers are traversed to find AI-detected targets that successfully match targets detected by RuleBase, and AI-detected targets with filter tags are deleted, including: traversing all and deleting AI-detected targets that match RuleBase and have filter tags.
[0078] By traversing all trackers, find the AI detection target that matches the RuleBase detection successfully, then check whether the target has a filter tag. If so, remove the target. If not, no further operation is required.
[0079] The present application embodiment also provides a false detection processing device 200, such as Figure 2 As shown, a schematic diagram of the structure of the false detection processing device in an embodiment of the present application is provided, and the false detection processing device 200 at least includes: an acquisition module 210, a tracker matching module 220, a feature calculation module 230 and a traversal elimination module 240, wherein:
[0080] In one embodiment of the present application, the acquisition module 210 is specifically used to: acquire the target of AI detection and the target of RuleBase detection, the AI detection includes an artificial intelligence detection method, and the RuleBase detection includes a rule-based detection method.
[0081] AI detection is an artificial intelligence detection method that can detect and identify targets and their types. RuleBase is a rule-based detection method that can be used as a fallback solution when AI detection may have incomplete data sets and cannot accurately identify targets. The target type detected by RuleBase is unknown.
[0082] The parallel architecture of AI-detected targets and RuleBase-detected targets can avoid accidental deletion of targets or scenarios where AI detection cannot identify them.
[0083] In one embodiment of the present application, the tracker matching module 220 is specifically used to: perform front and back frame matching in the tracker and obtain a tracker matching result based on the target detected by the AI and the target detected by the RuleBase, and the tracker matching result includes the AI tracker and RuleBase tracker corresponding to the real target.
[0084] According to the pre-input of the target detected by AI and the target detected by RuleBase, the previous and next frames are tracked and matched to obtain the tracker matching result. The tracker matching result includes the AI tracker and RuleBase tracker corresponding to the real target, and the feature information related to the detected target is also synchronously put into the tracker to obtain the RuleBase target that matches the AI detection.
[0085] In one embodiment of the present application, the feature calculation module 230 is specifically used to: calculate the features of the RuleBase tracker that matches the AI tracker in the tracker matching results, and add a filtering tag to the AI detected target whose features meet preset conditions.
[0086] Through the constructed tracker, the targets can be classified and matched with each other to obtain the AI tracker and RuleBase tracker corresponding to the real target.
[0087] By further calculating the features of the RuleBase tracker that matches the AI tracker in the tracker matching results, a filtering label is added to the AI target corresponding to RuleBase.
[0088] In one embodiment of the present application, the traversal elimination module 240 is specifically used for:
[0089] Traverse all trackers, find the AI-detected target that matches the target detected by RuleBase successfully, and delete the AI-detected target with the filter label.
[0090] Considering the possible false detection, all trackers are traversed to find the corresponding AI detection target when AI detection and RuleBase detection match successfully, and then the target is removed according to the filtering label to complete the false detection processing. The vehicle environment perception after false detection processing is input into the downstream decision planning.
[0091] In one embodiment of the present application, the feature calculation module 230 is used to
[0092] Calculate the height value of the RuleBase detection target based on the center point and size of the RuleBase detection target;
[0093] Calculate the angle value of the RuleBase detection target according to the direction of the RuleBase detection target, where the angle value is the angle from the positive x-axis of the vehicle to the vector counterclockwise;
[0094] According to the size of the RuleBase detection target matched by AI detection, calculate the aspect ratio of the RuleBase detection target.
[0095] In one embodiment of the present application, the feature calculation module 230 is also used to
[0096] If the height value of the RuleBase detection target is greater than the preset threshold value, the angle value is within the angle setting range and the aspect ratio is greater than the set threshold, the filtering label is added to the AI detection target corresponding to the RuleBase detection target, and the preset threshold value, the angle setting range and the set threshold are prior knowledge.
[0097] In one embodiment of the present application, the acquisition module is also used to
[0098] Get the size, center point, direction and type of the target detected by AI;
[0099] Get the size, center point, direction, and type of the target detected by RuleBase.
[0100] In one embodiment of the present application, the tracker matching module 220 is also used to
[0101] Put the target detected by the AI and the target detected by the RuleBase into a tracking and matching module for processing, and the tracking and matching module is used to track and match the previous and next frames;
[0102] In the tracking and matching module, the target input in the first frame is used as a new target and a corresponding target tracker is generated;
[0103] Match each frame target with the target tracker and create a new tracker for unmatched targets.
[0104] In one embodiment of the present application, the tracker matching module 220 is also used to
[0105] By determining whether the 3D rectangular frame of the target detected by AI intersects with the target contour frame detected by RuleBase;
[0106] If there is an intersection, the trackers are matched with each other and the successfully matched tracker is obtained;
[0107] Synchronize the target’s feature information to the successfully matched tracker.
[0108] In one embodiment of the present application, the traversal elimination module 240 is also used to
[0109] Traverse all trackers to find the corresponding AI detection targets when AI detection and RuleBase detection match successfully, and eliminate the targets of AI detection that match RuleBase and have filter labels.
[0110] It can be understood that the above-mentioned false detection processing device can implement each step of the false detection processing method provided in the above-mentioned embodiment, and the relevant explanations about the false detection processing method are applicable to the false detection processing device, which will not be repeated here.
[0111] Figure 3 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present application. Figure 3 At the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and a memory. The memory may include a memory, such as a high-speed random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage. Of course, the electronic device may also include hardware required for other services.
[0112] The processor, network interface and memory can be interconnected through an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.
[0113] The memory is used to store the program. Specifically, the program may include a program code, and the program code includes a computer operation instruction. The memory may include a memory and a non-volatile memory, and provides instructions and data to the processor.
[0114] The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming an error detection processing device at the logical level. The processor executes the program stored in the memory and is specifically used to perform the following operations:
[0115] Obtaining a target for AI detection and a target for RuleBase detection, wherein the AI detection includes an artificial intelligence detection method, and the RuleBase detection includes a rule-based detection method;
[0116] According to the target detected by the AI and the target detected by the RuleBase, matching is performed in the tracker between the previous and next frames and a tracker matching result is obtained, wherein the tracker matching result includes the AI tracker and the RuleBase tracker corresponding to the real target;
[0117] Calculating the features of the RuleBase tracker matched with the AI tracker in the tracker matching result, and adding a filter tag to the AI detection target whose features meet the preset conditions; and
[0118] Traverse all trackers, find the corresponding AI detection target when AI detection and RuleBase detection match successfully, and remove the target according to the filtering label.
[0119] The above application Figure 1 The method performed by the misdetection processing device disclosed in the illustrated embodiment can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or an instruction in the form of software. The above processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present application can be directly embodied as a hardware decoding processor for execution, or a combination of hardware and software modules in the decoding processor for execution. The software module can be located in a storage medium mature in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.
[0120] The electronic device may also perform Figure 1 The method is executed by the false detection processing device in Figure 1 The functions of the illustrated embodiment will not be described in detail in the embodiments of the present application.
[0121] The present application also provides a computer-readable storage medium, which stores one or more programs, wherein the one or more programs include instructions, which, when executed by an electronic device including multiple application programs, enable the electronic device to execute Figure 1 The method performed by the false detection processing device in the embodiment shown is specifically used to perform:
[0122] Obtaining a target for AI detection and a target for RuleBase detection, wherein the AI detection includes an artificial intelligence detection method, and the RuleBase detection includes a rule-based detection method;
[0123] According to the target detected by the AI and the target detected by the RuleBase, matching is performed in the tracker between the previous and next frames and a tracker matching result is obtained, wherein the tracker matching result includes the AI tracker and the RuleBase tracker corresponding to the real target;
[0124] Calculating the features of the RuleBase tracker matched with the AI tracker in the tracker matching result, and adding a filter tag to the AI detection target whose features meet the preset conditions; and
[0125] Traverse all trackers, find the corresponding AI detection target when AI detection and RuleBase detection match successfully, and remove the target according to the filtering label.
[0126] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0127] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0128] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0129] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.
[0130] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0131] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.
[0132] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.
[0133] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.
[0134] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.
[0135] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.
Claims
1. A method for handling false positives, wherein: The method comprises: Obtaining a target for AI detection and a target for RuleBase detection, wherein the AI detection includes an artificial intelligence detection method, and the RuleBase detection includes a rule-based detection method; According to the target detected by the AI and the target detected by the RuleBase, matching is performed in the tracker between the previous and next frames and a tracker matching result is obtained, wherein the tracker matching result includes the AI tracker and the RuleBase tracker corresponding to the real target; Calculating the features of the RuleBase tracker matched with the AI tracker in the tracker matching result, and adding a filter tag to the AI detection target whose features meet the preset conditions; and Traverse all trackers, find AI-detected targets that successfully match those detected by RuleBase, and delete AI-detected targets with filtered labels.
2. The method of claim 1, wherein: The calculating the feature of the RuleBase tracker matched with the AI tracker in the tracker matching result includes: Calculate the height value of the RuleBase detection target based on the center point and size of the RuleBase detection target; Calculate the angle value of the RuleBase detection target according to the direction of the RuleBase detection target, where the angle value is the angle of the vector rotated counterclockwise from the positive x-axis direction of the vehicle; According to the size of the RuleBase detection target, calculate the aspect ratio of the RuleBase detection target.
3. The method of claim 2, wherein: The adding of a filter label to an AI-detected target whose features meet preset conditions includes: If the height value of the target detected by the RuleBase is greater than the preset threshold value, the angle value is within the angle setting range, and the aspect ratio is greater than the set threshold, the filtering label is added to the AI detected target corresponding to the target detected by the RuleBase, and the preset threshold value, the angle setting range and the set threshold are prior knowledge.
4. The method of claim 1, wherein: The acquisition of the AI detection target and the RuleBase detection target includes: Get the target detected by AI, including the size, center point, direction and type of the target; Get the target detected by RuleBase, including the size, center point, direction, and type of the target.
5. The method of claim 1, wherein: According to the target detected by the AI and the target detected by the RuleBase, matching the previous and next frames in the tracker and obtaining the tracker matching result includes: Put the target detected by the AI and the target detected by the RuleBase into a tracking and matching module for processing, and the tracking and matching module is used to track and match the previous and next frames; In the tracking and matching module, the target input in the first frame is used as a new target and a corresponding target tracker is generated; Match each frame target with the target tracker and create a new tracker for unmatched targets.
6. The method of claim 1, wherein: According to the target detected by the AI and the target detected by the RuleBase, matching the previous and next frames in the tracker and obtaining the tracker matching result includes: Determine whether the 3D rectangular box of the target detected by AI intersects with the contour box of the target detected by RuleBase; If there is an intersection, the trackers are matched with each other and the successfully matched tracker is obtained; Synchronize the target’s feature information to the successfully matched tracker.
7. The method of claim 1, wherein: Traverse all trackers, find AI-detected targets that successfully match those detected by RuleBase, and delete AI-detected targets with filter tags, including: Traverse all trackers to find the corresponding AI detection target when AI detection and RuleBase detection match successfully, and delete the AI detection target that matches RuleBase and has a filter tag.
8. A false detection processing device, wherein: The device comprises: An acquisition module, used to acquire a target for AI detection and a target for RuleBase detection, wherein the AI detection includes an artificial intelligence detection method, and the RuleBase detection includes a rule-based detection method; A tracker matching module, used to match the previous and next frames in the tracker according to the target detected by the AI and the target detected by the RuleBase, and obtain a tracker matching result, wherein the tracker matching result includes the AI tracker and the RuleBase tracker corresponding to the real target; a feature calculation module, used to calculate the features of the RuleBase tracker matched with the AI tracker in the tracker matching result, and add a filter tag to the AI detection target whose features meet the preset conditions; and The traversal and elimination module is used to traverse all trackers, find the AI detected targets that match the targets detected by RuleBase successfully, and delete the AI detected targets with filter tags.
9. An electronic device, comprising: processor; as well as A memory arranged to store computer executable instructions, which when executed cause the processor to perform the method of any one of claims 1 to 7.
10. A computer-readable storage medium storing one or more programs, which, when executed by an electronic device including a plurality of application programs, causes the electronic device to execute any one of the methods of claims 1 to 7.