Incomplete target arbitration method and device of vehicle, vehicle and storage medium

Abstract

The application relates to the technical field of target detection, in particular to a vehicle incomplete target arbitration method and device, a vehicle and a storage medium, wherein the method comprises the following steps: identifying multiple targets of a vehicle to generate a target set; constructing a detection subgraph according to the target set and acquiring an isomorphic subgraph of the detection subgraph in a preset reference graph, wherein the preset reference graph is constructed by a reference image obtained by global target labeling; and arbitrating the predicted category of each node in the detection subgraph by using the isomorphic subgraph to obtain a final determination result. According to the vehicle incomplete target arbitration method, prior information of a target image is introduced in target detection, and the target is arbitrated to determine the target category, so that the accuracy of target detection is improved.

Description

Technical Field

[0001] This application relates to the field of target detection technology, and in particular to a method, apparatus, vehicle, and storage medium for arbitrating incomplete targets of a vehicle. Background Technology

[0002] In recent years, with the rise of deep learning technology, object detection has made rapid progress in the field of computer vision and has been widely used in fields such as face recognition and intelligent driving.

[0003] In target detection technology, there are usually two situations that lead to incomplete target acquisition: one is that the target object is partially obscured by obstacles, and the other is that the target object is captured at the edge, which will cause the captured image features to be incomplete or disappear. In this case, if the performance of the detector is aging, it is easy to lead to inaccurate recognition results and missed recall. Moreover, when the target is incomplete, the detection difficulty from a single static image is much greater than the detection difficulty from a dynamic image.

[0004] In related technologies, the problem of incomplete target recognition in static scenes is mainly addressed by utilizing the pixel information of incomplete targets to optimize the detection algorithm, and by training algorithm models with different degrees of occlusion to improve the generalization of the model.

[0005] However, this method has limited ability to identify target images with a large degree of missing information and missing key information features, which increases the false recognition rate of target images. Summary of the Invention

[0006] This application provides a method, apparatus, vehicle, and storage medium for arbitrating incomplete targets in vehicles, in order to solve the problem that the recognition capability for target images with a large degree of missing parts is limited in static scenes, thus making it impossible to accurately detect incomplete targets.

[0007] The first aspect of this application provides a method for arbitrating incomplete targets of a vehicle, comprising the following steps: identifying multiple targets of the vehicle and generating a target set; constructing a detection subgraph based on the target set and obtaining a isomorphic subgraph of the detection subgraph in a preset benchmark graph, wherein the preset benchmark graph is constructed from a benchmark image obtained by global target annotation; and arbitrating the predicted category of each node in the detection subgraph using the isomorphic subgraph to obtain a final judgment result.

[0008] Based on the above technical means, the accuracy of target detection is improved by introducing prior information of the target image and arbitrating the target to determine the target category.

[0009] Further, in an embodiment of the present application, before obtaining the isomorphic subgraph of the detection subgraph in the preset reference graph, further comprising: collecting and labeling the reference image; calculating the center point of each labeled box and the distance between each center point according to the reference image, and normalizing to obtain distance data and node categories; constructing the preset reference graph according to the distance data and the node categories.

[0010] According to the above technical means, the reference graph is obtained by calculating the center point of each labeled box and the distance between each center point, thereby improving the position of the node in the reference image.

[0011] Further, in an embodiment of the present application, before calculating the center point of each labeled box and the distance between each center point according to the reference image, further comprising: identifying the current scene of the vehicle; matching the best distance calculation method according to the current scene, so as to calculate the center point of each labeled box and the distance between each center point according to the best distance calculation method.

[0012] According to the above technical means, by selecting the corresponding distance calculation formula according to different scenes, the calculation method is facilitated and the efficiency is improved.

[0013] Further, in an embodiment of the present application, the obtaining the isomorphic subgraph of the detection subgraph in the preset reference graph comprises: obtaining the isomorphic subgraph by using a preset subgraph matching method based on the mapping relationship between the detection subgraph and the preset reference graph.

[0014] According to the above technical means, the position accuracy of the detection subgraph in the reference graph is improved by the obtained isomorphic subgraph.

[0015] Further, in an embodiment of the present application, the using the isomorphic subgraph to arbitrate the predicted category of each node in the detection subgraph comprises: obtaining the reference category of each node according to the isomorphic subgraph; determining the reference category and the detection category by a preset arbitration method to obtain the final category of each node.

[0016] According to the above technical means, by arbitrating the predicted category of each node in the detection subgraph, the category and detection accuracy of the target image are improved.

[0017] Further, in an embodiment of the present application, the determining the final category of each node by the preset arbitration method includes: traversing the detection subgraph to obtain a predicted category and a confidence of the detection subgraph; determining whether the predicted category of the detection subgraph matches the reference category, and obtaining a correct node and a confused node according to the determination result; traversing the confused node, obtaining a node set connected to the confused node, and voting for the confused node to obtain a voting confidence of the confused node; determining whether the voting confidence is higher than a preset threshold, wherein if the voting confidence is higher than the preset threshold, the confused node is determined as the reference category, otherwise, the confused node is determined as the predicted category; and until all confused nodes are traversed to obtain the final determination result.

[0018] According to the above technical means, the category and detection accuracy of the target image are improved by arbitrating the predicted category of each node in the detection subgraph, so that the final category is obtained.

[0019] The second aspect embodiment of the present application provides an incomplete target arbitration device of a vehicle, which includes: an identification module configured to identify a plurality of targets of the vehicle to generate a target set; an acquisition module configured to construct a detection subgraph according to the target set, and acquire an isomorphic subgraph of the detection subgraph in a preset reference graph, wherein the preset reference graph is constructed by a reference image labeled by a global target; and an arbitration module configured to arbitrate a predicted category of each node in the detection subgraph by using the isomorphic subgraph to obtain a final determination result.

[0020] Further, in an embodiment of the present application, before acquiring the isomorphic subgraph of the detection subgraph in the preset reference graph, the acquisition module further includes: a collection unit configured to collect and label the reference image; a calculation unit configured to calculate a center point of each labeled frame and a distance between each center point according to the reference image, and perform normalization to obtain distance data and node categories; and a construction unit configured to construct the preset reference graph according to the distance data and the node categories.

[0021] Further, in an embodiment of the present application, before calculating the center point of each labeled frame and the distance between each center point according to the reference image, the calculation unit is further configured to: identify a current scene of the vehicle; and match a best distance calculation method according to the current scene to calculate the center point of each labeled frame and the distance between each center point according to the best distance calculation method.

[0022] Further, in an embodiment of the present application, the obtaining module comprises: a first obtaining unit, configured to obtain the isomorphic subgraph based on the mapping relationship between the detection subgraph and the preset reference graph by using a preset subgraph matching method.

[0023] Further, in an embodiment of the present application, the arbitration module comprises: a second obtaining unit, configured to obtain the reference category of each node according to the isomorphic subgraph; and a judging unit, configured to judge the reference category and the detection category by using a preset arbitration method to obtain the final category of each node.

[0024] Further, in an embodiment of the present application, the judging unit is specifically configured to: traverse the detection subgraph to obtain a predicted category and a confidence of the detection subgraph; judge whether the predicted category and the reference category of the detection subgraph are matched, and determine a correct node and an ambiguous node according to a judgment result; traverse the ambiguous node, obtain a node set connected with the ambiguous node, and vote for the ambiguous node to obtain a voting confidence of the ambiguous node; judge whether the voting confidence is higher than a preset threshold, wherein if the voting confidence is higher than the preset threshold, the ambiguous node is judged as the reference category, otherwise, the ambiguous node is judged as the predicted category; and traverse all ambiguous nodes until a final judgment result is obtained.

[0025] A third aspect of the present application provides a vehicle, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the incomplete target arbitration method of the vehicle as described in the above embodiments.

[0026] A fourth aspect of the present application provides a computer readable storage medium having a computer program stored thereon, wherein the program is executed by a processor to implement the incomplete target arbitration method of the vehicle as described in the above embodiments.

[0027] The embodiments of the present application identify multiple targets of a vehicle, generate a target set, construct a detection subgraph according to the target set, obtain an isomorphic subgraph of the detection subgraph in a preset reference graph, and arbitrate a predicted category of each node in the detection subgraph by using the isomorphic subgraph to obtain a final judgment result. Thus, the problem that the recognition ability of a target image with a large degree of loss is limited in a static scene, so that an incomplete target cannot be accurately detected is solved.

[0028] Additional aspects and advantages of the present application will be in part apparent and in part pointed out hereinafter. BRIEF DESCRIPTION OF DRAWINGS

[0029] The above and / or additional aspects and advantages of the present application will become apparent and more readily appreciated from the following description, taken in conjunction with the following drawings of which:

[0030] Figure 1 A flowchart of an incomplete target arbitration method for a vehicle according to an embodiment of the present application;

[0031] Figure 2 A flowchart of target detection based on prior information according to an embodiment of the present application;

[0032] Figure 3 A schematic diagram of detecting confusing targets in a faulty light recognition scenario according to an embodiment of the present application;

[0033] Figure 4 A schematic diagram of constructing a reference map according to relative positions of each target in a reference image according to an embodiment of the present application;

[0034] Figure 5 A flowchart of an arbitration method according to an embodiment of the present application;

[0035] Figure 6 An example diagram of an incomplete target arbitration device for a vehicle according to an embodiment of the present application;

[0036] Figure 7 A structural schematic diagram of a vehicle according to an embodiment of the present application.

[0037] Reference Signs List: 10 - incomplete target arbitration device for a vehicle; 100 - identification module, 200 - acquisition module, 300 - arbitration module. DETAILED DESCRIPTION

[0038] Embodiments of the present application are described in detail below with reference to the attached drawing figures, wherein the same or like reference numerals and characters are used throughout the figures to denote same or like components. The embodiments described below are illustrative only, and are not intended to be limiting on the present application.

[0039] An incomplete target arbitration method, device, vehicle and storage medium of a vehicle are described below with reference to the accompanying drawings. In view of the limited recognition capability of a target image with a large degree of incompleteness in a static scene, which makes it impossible to accurately detect an incomplete target, the present application provides an incomplete target arbitration method of a vehicle. In the method, a plurality of targets of the vehicle are identified to generate a target set, and a detection subgraph is constructed according to the target set to obtain an isomorphic subgraph of the detection subgraph in a preset reference graph. Meanwhile, the isomorphic subgraph is used to arbitrate the predicted category of each node in the detection subgraph to obtain a final determination result. Thus, the problem of limited recognition capability of a target image with a large degree of incompleteness in a static scene, which makes it impossible to accurately detect an incomplete target, is solved. The prior information of a target image is introduced in target detection, and the target is arbitrated to determine the target category, thereby improving the accuracy of target detection.

[0040] Specifically, Figure 1 A flowchart of an incomplete target arbitration method of a vehicle provided by the present application is shown.

[0041] As Figure 1 shown, the incomplete target arbitration method of the vehicle includes the following steps:

[0042] In step S101, a plurality of targets of the vehicle are identified to generate a target set.

[0043] Specifically, as Figure 2 shown, in a static scene, the prior information of the target position information of the vehicle is fused into target detection, one or more targets in the target set contained in the prior information are identified, and the positions and categories of all to-be-detected targets in the static scene are labeled to obtain a reference image of the to-be-detected target, so as to determine whether there is a to-be-detected target, and accurately identify the category of the to-be-detected target through assistance.

[0044] In the present application, the prior information can be constructed in different scenes. One or more static target images in the target set can be collected and labeled. For example, when a single image is collected and labeled, the labeled information needs to contain the information of all to-be-detected targets. If multiple target images are collected and labeled, the information set after image labeling needs to cover the information of all to-be-detected targets and can be used to construct a reference graph.

[0045] For example, as Figure 3 shown, Figure 3For the five vehicle instrument panel fault light images to be detected, light up the entire instrument panel fault light, collect the complete instrument panel fault light image and label the fault light image to get the reference image, wherein the collected and labeled reference image includes the position data and category of all target to be detected. That is, if it is known in advance which fault light is at the c position in Figure 2 , the category of e can be accurately inferred by referring to the relative position of c and e.

[0046] In step S102, a detection subgraph is constructed according to the target set, and a homomorphic subgraph of the detection subgraph in a preset reference graph is obtained, wherein the preset reference graph is constructed by the reference image labeled by the global target.

[0047] Further, in an embodiment of the present application, before obtaining the homomorphic subgraph of the detection subgraph in the preset reference graph, it further includes: collecting and labeling the reference image; calculating the center point of each labeled frame and the distance between the center points according to the reference image, and normalizing to obtain the distance data and node category; and constructing the preset reference graph according to the distance data and node category.

[0048] Further, in an embodiment of the present application, before calculating the center point of each labeled frame and the distance between the center points according to the reference image, it further includes: identifying the current scene of the vehicle; and matching the best distance calculation method according to the current scene to calculate the center point of each labeled frame and the distance between the center points according to the best distance calculation method.

[0049] Specifically, the embodiments of the present application calculate the relative positions between categories according to the position data and category of the target to be detected labeled by the reference image, thereby obtaining the relative position data. Different distance calculation methods can be selected according to different scenes to calculate the center point of each labeled frame and the distance between the center points according to the best distance calculation method, for example, for the scene of fault light identification, a static plane image, Euler distance can be selected, which has the advantage of being not affected by image rotation and translation.

[0050] Specifically, the embodiments of the present application calculate the center points of each category according to the labeled frames of each category, and then calculate the center points of each category and the distance between the center points, and normalize the distance between each target category to obtain the absolute distance and node category, and convert the absolute distance to relative distance, thereby obtaining the relative distance data between each target category, and then constructing the preset reference graph according to the relative distance data and node category.

[0051] For example, as shown in Figure 4As shown, in the embodiment of the present application, the center points of all target categories to be detected are connected one by one to form a connected graph, in which the graph contains six rectangular frame corresponding six targets, such as A-F, the categories of the targets can be the same or different, and the targets are connected by straight lines, the straight line represents the distance, and the weight of the straight line represents the relative distance. Through the category and distance relationship between multiple target nodes, a certain node in the graph can be accurately positioned.

[0052] Further, after constructing the preset reference graph according to the distance data and the node category, the embodiment of the present application performs target detection on the preset reference graph by using a general target detection algorithm, such as YOLO (You Only Look Once, target detection algorithm), SSD (Single Shot MultiBox Detector, target detection algorithm)-H and R-CNN (Region-Convolutional Neural Networks, region convolutional neural network) series algorithm, thereby outputting a pre-recognized target set, wherein each target contains a category, a detection frame coordinate and a confidence. At the same time, for each target in the pre-recognized target set, the position data of each target category is repeatedly calculated, thereby obtaining a connected graph of the pre-recognized target set. It should be noted that the connected graph is essentially a subgraph of the reference graph, so the prediction of the target category at this time can be converted into judging the position of the detection subgraph in the reference graph, that is, a detection subgraph can be constructed according to the pre-recognition result.

[0053] Further, in an embodiment of the present application, obtaining the isomorphic subgraph of the detection subgraph in the preset reference graph comprises: obtaining the isomorphic subgraph by using a preset subgraph matching method based on the mapping relationship between the detection subgraph and the preset reference graph.

[0054] The preset subgraph matching method can be a matching method selected by a person skilled in the art, or a matching method obtained through computer simulation multiple times, which is not limited here.

[0055] Specifically, the embodiment of the present application completes the mapping of the detection subgraph to the preset reference graph based on the mapping relationship between the detection subgraph and the preset reference graph by using the preset subgraph matching method, thereby obtaining the isomorphic subgraph of the detection subgraph in the preset reference graph, and further obtaining the reference category of each node in the preset reference graph.

[0056] In step S103, the predicted category of each node in the detection subgraph is arbitrated by using the isomorphic subgraph to obtain a final determination result.

[0057] Further, in an embodiment of the present application, the arbitration of the isomorphic subgraph is used to detect the predicted category of each node in the detection subgraph, which includes: obtaining the reference category of each node according to the isomorphic subgraph; and determining the reference category and the detection category by using a preset arbitration method to obtain the final category of each node.

[0058] The preset arbitration method can be an arbitration method selected by a person skilled in the art, or an arbitration method obtained through computer simulation multiple times, and is not limited here

[0059] Specifically, after obtaining the reference category of each node according to the isomorphic subgraph of the detection subgraph in the preset reference graph, the reference category and the detection category are determined by using the preset arbitration method, so as to obtain the final category of each node.

[0060] Further, in an embodiment of the present application, the arbitration of the isomorphic subgraph is used to detect the predicted category of each node in the detection subgraph, which includes: obtaining the reference category of each node according to the isomorphic subgraph; and determining the reference category and the detection category by using a preset arbitration method to obtain the final category of each node.

[0061] Specifically, as shown in Figure 5 the arbitration process of the reference category and the detection category in the isomorphic subgraph, first, the detection subgraph is traversed to obtain the predicted category and the confidence of each detection subgraph, and it is respectively judged whether the reference category and the detection category of the detection subgraph are consistent. If they are consistent, the detection category is identified as correct, and the corresponding node is a correct node. If they are not consistent, the detection category is identified as incorrect, and the corresponding node is a confused node, and the confused node is added to the confused node set.

[0062] Secondly, each confused node is traversed to obtain a node set connected with the confused node, and the confused node is voted by the node set to judge whether the detection category is correct, so as to obtain the voting confidence of the confused node. It should be noted that the confused node used for voting can include a confused node that has not been arbitrated, or a confused node that is being arbitrated itself. At this time, a weighted average method is used to give a higher weight to the correct node and a lower weight to the confused node.

[0063] Furthermore, this embodiment requires setting a threshold for the voting confidence level to determine the category of the obfuscated node. For example, this embodiment can set the preset threshold to 0.8. When the voting confidence level is higher than 0.8, the category is determined as the baseline category; otherwise, it is determined as the detection category. At the same time, the node with the updated category is regarded as the correct node, and the average confidence level after voting is used as the node confidence level. This process continues until all obfuscated nodes are traversed to obtain the final determination result.

[0064] According to the incomplete target arbitration method for vehicles in this application, multiple targets of the vehicle are identified, a target set is generated, and a detection subgraph is constructed based on the target set. An isomorphic subgraph of the detection subgraph in a preset baseline map is obtained, and the predicted category of each node in the detection subgraph is arbitrated using the isomorphic subgraph to obtain the final judgment result. This solves the problem that in static scenes, the recognition capability for target images with significant missing parts is limited, thus hindering accurate detection of incomplete targets. By introducing prior information of the target image into target detection and arbitrating the target to determine the target category, the accuracy of target detection is improved.

[0065] Next, referring to the accompanying drawings, an incomplete target arbitration device for a vehicle proposed according to an embodiment of this application is described.

[0066] Figure 6 This is a block diagram of an incomplete target arbitration device for a vehicle according to an embodiment of this application.

[0067] like Figure 6 As shown, the incomplete target arbitration device 10 of the vehicle includes: an identification module 100, an acquisition module 200, and an arbitration module 300.

[0068] The identification module 100 is used to identify multiple targets on the vehicle and generate a target set.

[0069] The acquisition module 200 is used to construct a detection sub-map based on the target set and acquire isomorphic sub-maps of the detection sub-maps in a preset benchmark map, wherein the preset benchmark map is constructed from a benchmark image obtained by global target annotation; and

[0070] Arbitration module 300 is used to detect the predicted category of each node in the isomorphic subgraph by arbitrating the subgraph and obtain the final judgment result.

[0071] Furthermore, in one embodiment of this application, before acquiring the isomorphic subgraph of the detection subgraph in the preset reference graph, the acquisition module 200 further includes: an acquisition unit, a calculation unit, and a construction unit.

[0072] The acquisition unit is used to acquire and label the reference image.

[0073] The computing unit is configured to calculate each annotation box center point and the distance between each center point according to the reference image, and perform normalization to obtain distance data and node categories;

[0074] The constructing unit is configured to construct a preset reference map according to the distance data and the node categories.

[0075] Further, in an embodiment of the present application, before calculating each annotation box center point and the distance between each center point according to the reference image, the computing unit is further configured to:

[0076] identify a current scene of the vehicle;

[0077] match a best distance calculation mode according to the current scene, so as to calculate each annotation box center point and the distance between each center point according to the best distance calculation mode.

[0078] Further, in an embodiment of the present application, the acquisition module 200 comprises:

[0079] The first acquisition unit is configured to obtain an isomorphic subgraph based on a mapping relationship between the detection subgraph and the preset reference map by using a preset subgraph matching method.

[0080] Further, in an embodiment of the present application, the arbitration module 300 comprises a second acquisition unit and a determination unit.

[0081] The second acquisition unit is configured to obtain a reference category of each node according to the isomorphic subgraph;

[0082] The determination unit is configured to determine the reference category and the detection category by using a preset arbitration method to obtain a final category of each node.

[0083] Further, in an embodiment of the present application, the determination unit is specifically configured to:

[0084] traverse the detection subgraph to obtain a predicted category and a confidence of the detection subgraph;

[0085] determine whether the predicted category of the detection subgraph matches the reference category, and determine correct nodes and confused nodes according to a determination result;

[0086] traverse the confused nodes, obtain a node set connected to the confused nodes, and vote for the confused nodes to obtain a voting confidence of the confused nodes;

[0087] determine whether the voting confidence is higher than a preset threshold, wherein if the voting confidence is higher than the preset threshold, the confused nodes are determined as the reference category, otherwise, the confused nodes are determined as the predicted category;

[0088] until the traversal of all the confused nodes is completed, a final determination result is obtained.

[0089] According to the incomplete target arbitration device of the vehicle provided in the embodiments of the present application, a target set is generated by identifying multiple targets of the vehicle, a detection subgraph is constructed according to the target set, an isomorphic subgraph of the detection subgraph in a preset reference graph is obtained, and the predicted category of each node in the detection subgraph is arbitrated by using the isomorphic subgraph to obtain a final determination result. In this way, the problem that the recognition ability of a target image with a large degree of loss in a static scene is limited and thus the incomplete target cannot be accurately detected is solved, the prior information of the target image is introduced in target detection, the target is arbitrated to determine the target category, and the accuracy of target detection is improved.

[0090] Figure 7 A structural schematic diagram of a vehicle is provided in the embodiments of the present application. The vehicle can include:

[0091] The memory 701, the processor 702, and the computer program stored in the memory 701 and executable on the processor 702.

[0092] The processor 702 implements the incomplete target arbitration method of the vehicle provided in the above embodiments when executing the program.

[0093] Further, the vehicle further includes:

[0094] The communication interface 703 is used for communication between the memory 701 and the processor 702.

[0095] The memory 701 is used to store the computer program executable on the processor 702.

[0096] The memory 701 can include a high-speed RAM (Random Access Memory, random access memory) memory, and can also include a non-volatile memory, for example, at least one disk memory.

[0097] If the memory 701, the processor 702, and the communication interface 703 are independently implemented, the communication interface 703, the memory 701, and the processor 702 can be connected to each other through a bus and complete communication between each other. The bus can be an ISA (Industry Standard Architecture, industry standard architecture) bus, a PCI (Peripheral Component, peripheral component interconnect) bus, or an EISA (Extended Industry Standard Architecture, extended industry standard architecture) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, Figure 7 In the figure, only one thick line is used to represent, but it does not mean that there is only one bus or one type of bus.

[0098] Optionally, in a specific implementation, if the memory 701, processor 702, and communication interface 703 are integrated on a single chip, then the memory 701, processor 702, and communication interface 703 can communicate with each other through an internal interface.

[0099] The processor 702 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of this application.

[0100] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for arbitrating incomplete vehicle targets.

[0101] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0102] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0103] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0104] It should be understood that portions of the application can be implemented in hardware, software, firmware, or combinations thereof. In the above embodiments, the N steps or methods can be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. As such, if implemented in hardware, and in another embodiment, any of the following technologies, known in the art, or their combinations, can be used to implement: discrete logic circuitry having logic gates for implementing logic functions on data signals, application specific integrated circuits having appropriate combinational logic gates, programmable gate arrays, field programmable gate arrays, etc.

[0105] Those skilled in the art of the present technology can understand that all or part of the steps carried out by the above-mentioned embodiment methods can be completed by programs instructing related hardware, and the programs can be stored in a computer readable storage medium. When the program is executed, it includes one of the steps of the method embodiment or a combination thereof.

[0106] Although the embodiments of the present application have been shown and described above, it should be understood that the above-mentioned embodiments are exemplary and cannot be understood as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above-mentioned embodiments within the scope of the present application.

Claims

1. A method of incomplete target arbitration for a vehicle, characterized by, The method comprises the following steps: identifying a plurality of targets of a vehicle to generate a target set; constructing a detection subgraph according to the target set, and obtaining an isomorphic subgraph of the detection subgraph in a preset reference graph, wherein the preset reference graph is constructed by a reference image obtained by global target labeling; and using the isomorphic subgraph to arbitrate the predicted category of each node in the detection subgraph to obtain a final determination result; the arbitration of the predicted category of each node in the detection subgraph using the isomorphic subgraph comprises: obtaining the reference category of each node according to the isomorphic subgraph; determining the reference category and the detection category by a preset arbitration method to obtain the final category of each node; determining the final category of each node by determining the reference category and the detection category comprises: traversing the detection subgraph to obtain the predicted category and the confidence of the detection subgraph; judging whether the predicted category and the reference category of the detection subgraph are matched, and obtaining correct nodes and confused nodes according to the judgment result; traversing the confused nodes, obtaining a node set connected to the confused nodes, and voting for the confused nodes to obtain the voting confidence of the confused nodes; judging whether the voting confidence is higher than a preset threshold, wherein if the voting confidence is higher than the preset threshold, the confused nodes are determined as the reference category, otherwise as the predicted category; until the traversal of all confused nodes is completed, the final determination result is obtained.

2. The method of claim 1, wherein, Before obtaining the isomorphic subgraph of the detection subgraph in the preset reference graph, the method further comprises: collecting and labeling the reference image; calculating the center point of each labeled box and the distance between each center point according to the reference image, and normalizing to obtain distance data and node categories; constructing the preset reference graph according to the distance data and the node categories.

3. The method of claim 2, wherein, Before calculating the center point of each labeled box and the distance between each center point according to the reference image, the method further comprises: identifying the current scene of the vehicle; matching the best distance calculation method according to the current scene to calculate the center point of each labeled box and the distance between each center point according to the best distance calculation method.

4. The method of claim 1, wherein, The method of obtaining the isomorphic subgraph of the detection subgraph in the preset reference graph comprises: using a preset subgraph matching method to obtain the isomorphic subgraph based on the mapping relationship between the detection subgraph and the preset reference graph.

5. An incomplete target arbitration apparatus for a vehicle, characterized by, comprises: an identification module configured to identify a plurality of targets of a vehicle to generate a target set; an acquisition module configured to construct a detection subgraph according to the target set, and obtain an isomorphic subgraph of the detection subgraph in a preset reference graph, wherein the preset reference graph is constructed by a reference image obtained by global target labeling; and an arbitration module configured to arbitrate the predicted category of each node in the detection subgraph using the isomorphic subgraph to obtain a final determination result; the arbitration module is specifically configured to obtain the reference category of each node according to the isomorphic subgraph; determine the final category of each node by determining the reference category and the detection category by a preset arbitration method; The arbitration module is further configured to traverse the detection subgraph to obtain a predicted category and a confidence of the detection subgraph; determine whether the predicted category and a reference category of the detection subgraph are matched, and determine a correct node and a confused node according to a determination result; traverse the confused node, obtain a node set connected to the confused node, and vote for the confused node to obtain a voting confidence of the confused node; determine whether the voting confidence is higher than a preset threshold, wherein if the voting confidence is higher than the preset threshold, the confused node is determined as the reference category, otherwise, the confused node is determined as the predicted category; traverse all confused nodes until a final determination result is obtained.

6. The apparatus of claim 5, wherein, Before obtaining the isomorphic subgraph of the detection subgraph in the preset reference graph, the obtaining module further includes: an acquisition unit configured to acquire and label the reference image; a calculation unit configured to calculate a center point of each labeled frame and a distance between each center point according to the reference image, and perform normalization to obtain distance data and node categories; a construction unit configured to construct the preset reference graph according to the distance data and the node categories.

7. A vehicle characterized by comprising: comprise: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to implement the incomplete target arbitration method of the vehicle according to any one of claims 1-4.

8. A computer readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the incomplete target arbitration method of the vehicle according to any one of claims 1-4.

Images