Label marking method, label marking device, medium and electronic equipment

Abstract

The present invention provides a labeling method, a labeling device, a medium and an electronic device. The labeling method comprises: obtaining a first vehicle image and processing the first vehicle image to obtain a vehicle track and a second vehicle image; classifying the second vehicle image to obtain a first vehicle image set; obtaining a first task file based on the first vehicle image set; screening the first task file to obtain a second task file; merging the second task file to associate task identifiers of the same vehicle. The labeling method can improve the efficiency of labeling vehicle images.

Description

Technical Field

[0001] The present invention relates to a vehicle data set labeling method, and in particular to a label labeling method, a label labeling device, a medium and an electronic device. Background Art

[0002] Vehicle re-identification technology is a hot technical field in computer vision technology and has high application value in vehicle-related fields. Since vehicle re-identification technology is restricted by the quantity and quality of training data sets, massive and high-quality annotated data sets are required to improve the accuracy of vehicle re-identification algorithms. In the data annotation process of vehicle re-identification, since the newly added data every day often has a correlation with the data of the previous few days, it is necessary to compare the data of the previous few days to establish an association relationship for the matching vehicles while annotating the newly added data. Therefore, the vehicle data set can generally only be annotated through inefficient manual annotation methods. Summary of the invention

[0003] In view of the above-mentioned shortcomings of the prior art, an object of the present invention is to provide a label marking method, a label marking device, a medium and an electronic device, so as to solve the problem of low efficiency of the manual marking method in the prior art.

[0004] To achieve the above-mentioned purpose and other related purposes, the first aspect of the present invention provides a labeling method, characterized in that the labeling method includes: acquiring a first vehicle image and processing the first vehicle image to obtain a vehicle trajectory and a second vehicle image, the second vehicle image is an image bound to the vehicle trajectory; classifying the second vehicle image to obtain a first vehicle image set, the elements in the first vehicle image set are image sets of the same vehicle; acquiring a first task file based on the first vehicle image set, the first task file includes a primary image and a secondary image, and each element in the first vehicle image set has a unique task identifier; screening the first task file to obtain a second task file; merging the second task file to associate task identifiers of the same vehicle.

[0005] In an embodiment of the first aspect, it also includes: processing the second vehicle image to obtain the structural similarity of adjacent images in the second vehicle image; judging whether the structural similarity of the adjacent images is greater than a first threshold, and if the structural similarity of the adjacent images is greater than the first threshold, deleting the image with the later time sequence in the adjacent images.

[0006] In an embodiment of the first aspect, a method for classifying the second vehicle images to obtain a first vehicle image set includes: extracting features of the second vehicle images through a neural network model to obtain similarities between the second vehicle images; and classifying the second vehicle images based on the similarities between the second vehicle images to obtain a first vehicle image set.

[0007] In an embodiment of the first aspect, it also includes: obtaining the intra-class similarity of the elements of the first vehicle image set; judging whether the intra-class similarity is greater than a second threshold or whether the intra-class similarity is less than a third threshold, and if the intra-class similarity is greater than the second threshold or the intra-class similarity is less than the third threshold, filtering the elements in the first vehicle image set.

[0008] In an embodiment of the first aspect, an implementation method for obtaining a first task file based on the first vehicle image set includes: generating a unique task identifier for an element in the first vehicle image set; obtaining pairwise similarities between images in the first vehicle image set; processing the first vehicle image set based on the pairwise similarities between the images to obtain the primary image and the secondary image, the primary image being associated with a unique task identifier of an element of the first vehicle image set corresponding to the primary image, and the secondary image being associated with a unique task identifier of an element of the first vehicle image set corresponding to the secondary image; and obtaining the first task file based on the primary image and the secondary image.

[0009] In an embodiment of the first aspect, the implementation method for obtaining the second task file includes: a first display step: using a first display device to display a first save button, a first previous task button, a first next task button, a first jump button, first task information, and a primary image and a secondary image corresponding to a current task identifier, the primary image corresponding to the current task identifier is displayed in a first image display area of ​​the first display device, the primary image corresponding to the current task identifier and the secondary image corresponding to the current task identifier are displayed in a second image display area of ​​the first display device, the first save button, the first previous task button, and the first next task button are displayed in a first function display area of ​​the first display device, and the first jump button and the first task information are displayed in a first task display area of ​​the first display device; a filtering step: filtering the images displayed in the first image display area and the second image display area according to filtering conditions; a first confirmation step: if the current task identifier is not the last task identifier, go to the filtering step, otherwise obtain the second task file.

[0010] In an embodiment of the first aspect, a method for implementing merging processing of the second task file includes: a second display step: using a second display device to display a second save button, a second previous task button, a second next task button, a second jump button, second task information, a primary image corresponding to a current task identifier, a primary image not associated with the current task identifier, and a primary image associated with the current task identifier, the second save button, the second previous task button, and the second next task button are displayed in a second function display area of ​​the second display device, the second jump button and the second task information are displayed in a second task display area of ​​the second display device, the primary image corresponding to the current task identifier is displayed in a third image display area of ​​the second display device, the primary image not associated with the current task identifier is displayed in a fourth image area of ​​the second display device, and the primary image associated with the current task identifier is displayed in a fifth image display area of ​​the second display device; an association step: performing association processing on the same vehicle image displayed in the third image display area and the fourth image display area to associate the task identifiers of the same vehicle image displayed in the third image display area and the fourth image display area; a second confirmation step: if the current task identifier is not the last task identifier, performing the association step based on the second function area.

[0011] A second aspect of the present invention provides a label marking device, including: a first vehicle image processing module, used to acquire a first vehicle image and process the first vehicle image to acquire a vehicle trajectory and a second vehicle image, the second vehicle image is an image bound to the vehicle trajectory; a first vehicle image set acquisition module, used to classify the second vehicle image to acquire a first vehicle image set, the elements in the first vehicle image set are image sets of the same vehicle; a first task file acquisition module, used to acquire a first task file based on the first vehicle image set, the first task file includes a primary image and a secondary image, and each element in the first vehicle image set has a unique task identifier; a second task file acquisition module, used to screen the first task file to acquire a second task file; a second task file processing module, used to merge the second task file to associate task identifiers of the same vehicle.

[0012] A third aspect of the present invention provides a computer-readable storage medium, and when the computer program is executed by a processor, the label marking method according to any one of the first aspects of the present invention is implemented.

[0013] The fourth aspect of the present invention provides an electronic device, comprising: a memory storing a computer program; a processor communicatively connected to the memory, for executing the label marking method described in any one of the first aspects of the present invention when the computer program is called; and a display communicatively connected to the processor and the memory, for displaying a GUI interactive interface related to the label marking method.

[0014] As described above, the label marking method, label marking device, medium and electronic device of the present invention have the following beneficial effects:

[0015] The labeling method includes classifying the acquired second vehicle image to obtain a first vehicle image set, wherein the elements in the first vehicle image set are image sets of the same vehicle; acquiring a first task file based on the first vehicle image set, wherein each element in the first vehicle image set has a unique task identifier; screening the first task file to obtain a second task file; and merging the second task file to associate the task identifiers of the same vehicle. The labeling method associates the unique task identifiers of the same vehicle through merging, thereby establishing an association relationship between the acquired new vehicle image and the old vehicle image, so the labeling method can improve the efficiency of labeling vehicle images. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Shown is a flow chart of a labeling method according to the present invention in a specific embodiment.

[0017] Figure 2 Shown is a flow chart of a labeling method according to the present invention in a specific embodiment.

[0018] Figure 3 Shown is a flow chart of a method for acquiring a first vehicle image set in a specific embodiment of the present invention.

[0019] Figure 4 Shown is a flow chart of a labeling method according to the present invention in a specific embodiment.

[0020] Figure 5 Shown is a flow chart of a method for implementing the method of acquiring the first task file in a specific embodiment of the present invention.

[0021] Figure 6 Shown is a flow chart of a method for implementing the method of acquiring the second task file in a specific embodiment of the present invention.

[0022] Figure 7 Shown is a schematic diagram of a display interface of a first display device in a specific embodiment of the label marking method of the present invention.

[0023] Figure 8 Shown is a flow chart of a specific embodiment of a method for implementing the merging process of the second task file according to the present invention.

[0024] Fig. 9 Shown is a schematic diagram of a display interface of a second display device in a specific embodiment of the labeling method of the present invention.

[0025] Fig.10 Shown is a schematic structural diagram of the label marking device of the present invention in a specific embodiment.

[0026] Fig.11 Shown is a schematic structural diagram of the electronic device of the present invention in a specific embodiment.

[0027] Component number description

[0028] 700 First display device

[0029] 710 First image display area

[0030] 720 Second image display area

[0031] 730 First function display area

[0032] 740 First Mission Display Area

[0033] 900 Second display device

[0034] 910 Third image display area

[0035] 920 Fifth image display area

[0036] 930 Second function display area

[0037] 940 Second Mission Display Area

[0038] 950 Fourth image display area

[0039] 1000 Labeling device

[0040] 1010 First vehicle image processing module

[0041] 1020 First vehicle image set acquisition module

[0042] 1030 First Task File Acquisition Module

[0043] 1040 Second Task File Acquisition Module

[0044] 1050 Second task file processing module

[0045] 1100 Electronic equipment

[0046] 1110 Memory

[0047] 1120 processor

[0048] 1130 Display

[0049] Steps S11-S15

[0050] Steps S21-S22

[0051] S31-S32 Steps

[0052] Steps S41-S42

[0053] Steps S51-S54

[0054] Steps S61-S63

[0055] Steps S81-S83 DETAILED DESCRIPTION

[0056] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0057] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and thus the drawings only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0058] In the data annotation process of vehicle re-identification, since the newly added data every day often has an association relationship with the data of the previous few days, it is necessary to compare the data of the previous few days to establish an association relationship with the matching vehicles while annotating the newly added data. Therefore, people can generally only annotate the vehicle data set through inefficient manual annotation methods. At least in view of the above problems, the present invention provides a label annotation method, which includes classifying the acquired second vehicle image to obtain a first vehicle image set, and the elements in the first vehicle image set are image sets of the same vehicle; based on the first vehicle image set, a first task file is acquired, and each element in the first vehicle image set has a unique task identifier; the first task file is screened to obtain a second task file; and the second task file is merged to associate the task identifier of the same vehicle. The label annotation method associates the unique task identifier of the same vehicle through merging, thereby establishing an association relationship between the acquired new vehicle image and the old vehicle image, so the label annotation method can improve the annotation efficiency of the vehicle image.

[0059] In one embodiment of the present invention, please refer to Figure 1 , the label marking method includes:

[0060] S11, acquiring a first vehicle image and processing the first vehicle image to acquire a vehicle track and a second vehicle image, wherein the second vehicle image is an image bound to the vehicle track, but the present embodiment is not limited thereto. The vehicle track may be a motion track of the vehicle, and the image bound to the vehicle track may be a captured image of the vehicle on the motion track of the vehicle.

[0061] Optionally, the implementation method of processing the first vehicle image to obtain a vehicle trajectory may be: processing the first vehicle image through a target detection algorithm to obtain a vehicle frame set of the first vehicle image; updating the vehicle trajectory based on the vehicle frame set, if the target vehicle on the vehicle trajectory is not updated, predicting the position information of the target vehicle through a Kalman filter tracking algorithm, if the target vehicle on the vehicle trajectory is updated by the vehicle frame set for a long time, deleting the target vehicle from the vehicle trajectory, and if the target vehicle on the vehicle trajectory is updated by the vehicle frame set for a short time, updating the vehicle trajectory based on the vehicle frame set.

[0062] Optionally, one implementation of the target vehicle on the vehicle trajectory being updated by the vehicle frame set for a long time may be: setting an update threshold for the target vehicle on the vehicle trajectory; if the number of times the target vehicle on the vehicle trajectory is updated by the vehicle frame set is less than the update threshold, the target vehicle on the vehicle trajectory is updated by the vehicle frame set for a long time. Similarly, if the number of times the target vehicle on the vehicle trajectory is updated by the vehicle frame set is greater than the update threshold, the target vehicle on the vehicle trajectory is updated by the vehicle frame set for a short time.

[0063] Optionally, the second vehicle image may include a captured image bound to the vehicle trajectory at the initial position of the vehicle trajectory, a captured image bound to the vehicle trajectory at the end position of the vehicle trajectory, and a captured image bound to the vehicle trajectory at a certain moment in the middle of the vehicle trajectory. The captured image bound to the vehicle trajectory at a certain moment in the middle of the vehicle trajectory may be a captured image bound to the vehicle trajectory retained at a fixed frame interval, for example, the captured image bound to the vehicle trajectory may be selected to be retained at a fixed interval of 25 frames.

[0064] Optionally, the labeling method may further include: processing the second vehicle image to obtain structural similarity of adjacent images in the second vehicle image; determining whether the structural similarity of the adjacent images is greater than a first threshold, and if the structural similarity of the adjacent images is greater than the first threshold, deleting the image that is earlier in time order in the adjacent images. The structural similarity is an indicator for measuring the similarity between two images.

[0065] S12, classify the second vehicle images to obtain a first vehicle image set, wherein the elements in the first vehicle image set are image sets of the same vehicle. Optionally, the implementation method of classifying the second vehicle images to obtain the first vehicle image set may be: processing the second vehicle images to obtain the pairwise similarities of the second vehicle images; classifying the second vehicle images based on the pairwise similarities of the second vehicle images to obtain the first vehicle image set, wherein the pairwise similarities of the second vehicle images may be the similarity between any two images in the second vehicle images, and the first vehicle image set may be, for example, {A, B, C}, wherein element A represents the image set of vehicle a, element B represents the image set of vehicle b, and element C represents the image set of vehicle c.

[0066] S13, based on the first vehicle image set, a first task file is obtained, wherein the first task file includes a primary image and a secondary image, and each element in the first vehicle image set has a unique task identifier. The primary image may be an image with the highest average similarity to other images in the image set of the same vehicle, and the secondary image may be an image remaining in the image set of the same vehicle except the primary image. The method for obtaining the average similarity may be, for example, that if the pairwise similarities between a picture and the other three pictures are 30, 40, and 50, respectively, then the average similarity between the picture and the other three pictures is 40.

[0067] Optionally, a unique random number may be generated by a random function; and the unique random number is set as a unique task identifier of an element in the first vehicle image set, but this embodiment is not limited thereto.

[0068] S14, filtering the first task file to obtain a second task file. Optionally, the method for filtering the first task file to obtain the second task file may include: reading the first task file; receiving a user's operation instruction to filter the first task file; and saving the filtered first task file to obtain the second task file.

[0069] S15, merging the second task files to associate task identifiers of the same vehicle. Optionally, the method for merging the second task files to associate task identifiers of the same vehicle may include: reading the second task files; merging the second task files to associate task identifiers of the same vehicle.

[0070] According to the above description, the labeling method described in this embodiment includes classifying the acquired second vehicle image to obtain a first vehicle image set, wherein the elements in the first vehicle image set are image sets of the same vehicle; obtaining a first task file based on the first vehicle image set, wherein each element in the first vehicle image set has a unique task identifier; screening the first task file to obtain a second task file; and merging the second task file to associate the task identifiers of the same vehicle. The labeling method associates the unique task identifiers of the same vehicle through merging, thereby establishing an association relationship between the acquired new vehicle image and the old vehicle image, so the labeling method can improve the efficiency of labeling vehicle images.

[0071] See also Figure 2 In one embodiment of the present invention, the label marking method further includes:

[0072] S21, processing the second vehicle image to obtain structural similarity of adjacent images in the second vehicle image, wherein the structural similarity is an indicator for measuring the similarity between two images.

[0073] S22, determining whether the structural similarity of the adjacent images is greater than a first threshold, and if the structural similarity of the adjacent images is greater than the first threshold, deleting the image that is later in time order among the adjacent images. The first threshold may be selected as 0.5, but the present invention is not limited thereto.

[0074] Optionally, the labeling method also includes processing the second vehicle image to obtain the clarity of the second vehicle image and the brightness of the second vehicle image; deleting images in the second vehicle image whose clarity is less than a clarity threshold or deleting images in the second vehicle image whose brightness is less than a brightness threshold, wherein the clarity threshold can be 0.5 and the brightness threshold can be 0.5, but this embodiment is not limited to this.

[0075] According to the above description, the labeling method described in this embodiment can screen the second vehicle image according to the structural similarity of adjacent images in the second vehicle image, which is conducive to improving the quality of the second vehicle image. In addition, the labeling method described in this embodiment can further improve the quality of the second vehicle image by deleting images with lower clarity or brightness.

[0076] See also Figure 3 In one embodiment of the present invention, a method for performing classification processing on the second vehicle image to obtain a first vehicle image set includes:

[0077] S31, extracting features from the second vehicle images through a neural network model to obtain similarities between the second vehicle images. The similarities between the second vehicle images may be similarities between two images in the second vehicle images. Optionally, the second vehicle images may be extracted features through a trained person re-identification model to obtain similarities between two images in the second vehicle images.

[0078] S32, classify the second vehicle images based on the similarity between the second vehicle images to obtain a first vehicle image set. Optionally, the second vehicle images may be processed by a clustering algorithm based on the similarity between the second vehicle images to obtain the first vehicle image set. The first vehicle image set may be, for example, {A, B, C}, where element A represents an image set of vehicle a, element B represents an image set of vehicle b, and element C represents an image set of vehicle c.

[0079] According to the above description, the labeling method described in this embodiment includes: classifying the second vehicle images based on the similarity between the second vehicle images to obtain a first vehicle image set. Since classified vehicle images are easier to label than discrete vehicle images, the labeling method described in this embodiment can effectively reduce the difficulty of labeling the second vehicle images.

[0080] See also Figure 4 In one embodiment of the present invention, the label marking method further includes:

[0081] S41, obtaining the intra-class similarity of the first vehicle image set elements. The intra-class similarity of the first vehicle image set elements may be the mean of the pairwise similarities of the images in the first vehicle image set elements. For example, in a vehicle set G, the vehicle set G includes three elements g1, g2, and g3, g1 includes three captured images p1, p2, and p3, d12 is the similarity between p1 and p2, d13 is the similarity between p1 and p3, d23 is the similarity between p2 and p3, and the intra-class similarity of the element g1 in the vehicle set G is the mean of d12, d13, and d23.

[0082] S42, determining whether the intra-class similarity is greater than a second threshold or less than a third threshold, and if the intra-class similarity is greater than the second threshold or less than the third threshold, filtering the elements in the first vehicle image set. The second threshold may be 0.9, and the third threshold may be 0.2, but the present embodiment is not limited thereto.

[0083] According to the above description, the labeling method described in this embodiment can effectively clean the dirty data in the first vehicle image set by obtaining the intra-class similarity of the elements of the first vehicle image set and filtering the elements in the first vehicle image set based on the intra-class similarity, thereby improving the cleaning efficiency and the quality of the first vehicle image set.

[0084] See also Figure 5 In one embodiment of the present invention, a method for obtaining a first task file includes:

[0085] S51: Generate a unique task identifier for an element in the first vehicle image set.

[0086] Optionally, the task identifier may be set to two states: valid or invalid. The task identifier may be initialized to a valid state when the task identifier is generated. When the task identifier is initialized, there is no association between the task identifiers.

[0087] S52, obtaining the pairwise similarity between images in the first vehicle image set. Optionally, a unique image identifier and image address of the image in the first vehicle image set may be generated to facilitate storage and search of the images in the first vehicle image set. The pairwise similarity between images in the first vehicle image set may be the similarity between any two images in the first vehicle image set.

[0088] S53, based on the pairwise similarity between the images, the first vehicle image set is processed to obtain the primary image and the secondary image, and the primary image and the secondary image are associated with the unique task identifier of the first vehicle image set element corresponding to the primary image and the secondary image. Optionally, when the primary image and the secondary image belong to the elements of the first vehicle image set element, the first vehicle image set element is the first vehicle set element corresponding to the primary image and the secondary image. For example, the first vehicle image set element includes three images M, N, and K, M is the primary image, N and K are secondary images, M, N, and K belong to the first vehicle image set element, and M, N, and K are associated with the unique task identifier of the first vehicle image set element. Among them, the association between the image and the task identifier indicates that there is a connection between the image and the task identifier, and one image can only be associated with one task identifier. The primary image may be an image with the highest average similarity to other images in a set of images of the same vehicle, and the secondary image may be the remaining images in the set of images of the same vehicle except the primary image. The method for obtaining the average similarity may be, for example: if the pairwise similarities between a picture and the other three pictures are 30, 40, and 50 respectively, then the average similarity between the picture and the other three pictures is 40.

[0089] S54, acquiring the first task file based on the primary image and the secondary image. Optionally, an implementation method for acquiring the first task file based on the primary image and the secondary image includes: generating the first task file based on the primary image, the secondary image, the image path of the primary image, the image path of the secondary image, the association relationship between the primary image and the unique task identifier corresponding to the primary image, and the association relationship between the secondary image and the unique task identifier corresponding to the secondary image; acquiring the first task file.

[0090] According to the above description, the labeling method described in this embodiment includes: generating a unique task identifier for the elements in the first vehicle image set, and obtaining the first task file based on the primary image and the secondary image. Since the difficulty of traversing the first vehicle image set through the first task file is less than the difficulty of directly traversing the first vehicle image set, the labeling method can improve the labeling efficiency.

[0091] See also Figure 6 and Figure 7 In one embodiment of the present invention, the method for obtaining the second task file includes a first display step, a filtering step and a first confirmation step.

[0092] In the first display step, the first display device is used to display the first save button, the first previous task button, the first next task button, the first jump button, the first task information, and the primary image and secondary image corresponding to the current task identifier. The primary image corresponding to the current task identifier is displayed in the first image display area of ​​the first display device, the primary image corresponding to the current task identifier and the secondary image corresponding to the current task identifier are displayed in the second image display area of ​​the first display device, the first save button, the first previous task button, and the first next task button are displayed in the first function display area of ​​the first display device, and the first jump button and the first task information are displayed in the first task display area of ​​the first display device. The task information may be the current task progress, each task identifier may represent a task, all task identifiers represent the total task amount, and the task progress may be represented by the completed task amount and the total task amount.

[0093] Optionally, the task information and image information displayed by the first display device may be updated by receiving a user's operation instruction, the image information may include a primary image and a secondary image corresponding to a specified task identifier, and the task information may be a task represented by the specified task identifier. Optionally, the user's operation instruction may be an instruction related to the first jump button.

[0094] In the filtering step, the images displayed in the first image display area and the second image display area are filtered according to a filtering condition, wherein the filtering condition may be any one of the following: the image is a non-vehicle image; the image is an image with severe vehicle occlusion; the secondary image is not the same vehicle as the primary image.

[0095] Optionally, a method for implementing filtering processing on the images displayed in the first image display area and the second image display area includes: receiving an operation instruction from a user to filter processing the images displayed in the first image display area and the second image display area.

[0096] Optionally, the method for receiving a user's operation instruction to filter the images displayed in the first image display area and the second image display area includes: receiving a user's operation instruction; setting the images displayed in the first image display area and the second image display area to a second state. Optionally, the images displayed in the first image display area and the second image display area are initially in a first state, and when the images displayed in the first image display area and the second image display area change from the first state to the second state, the first image display area and the second image display area no longer display the image, and the first image display area and the second image display area display other images in the first state.

[0097] Optionally, when the primary image is set to the second state, the unique task identifier associated with the primary image is set to an invalid state, and when the secondary image is set to the second state, the unique task identifier associated with the secondary image is set to no association.

[0098] Optionally, another method for implementing filtering processing on images displayed in the first image display area and the second image display area includes: filtering processing on images that meet filtering conditions and are displayed in the first image display area and the second image display area.

[0099] In the first confirmation step, if the current task identifier is not the last task identifier, the process proceeds to the filtering step, otherwise the second task file is acquired.

[0100] Optionally, whether the current task identifier is the last task identifier can be determined by receiving an operation instruction from the user. If, after receiving the operation instruction from the user, the images displayed in the first image display area and the second image display area are not updated, the current task identifier is the last task identifier.

[0101] Optionally, the user's operation instruction may be an operation instruction related to the first task button.

[0102] Optionally, the task identification state, the association relationship between the primary image and the secondary image and the corresponding task identification can be saved by responding to a user's operation instruction, and the second task file can be acquired based on the task identification state, the primary image, the secondary image, and the association relationship between the primary image and the secondary image and the corresponding task identification. The user's operation instruction can be an operation instruction related to the first save button.

[0103] According to the above description, the method for obtaining the second task file described in this embodiment reduces the dirty data in the first task file by traversing the primary images and secondary images in the first task file and filtering the images that meet the filtering conditions, thereby improving the data quality of the first task file.

[0104] See also Figure 8 and Fig. 9 In one embodiment of the present invention, the method for implementing the merging process of the second task file includes a second display step, an association step and a second confirmation step.

[0105] In the second display step, a second display device is used to display a second save button, a second previous task button, a second next task button, a second jump button, second task information, a primary image corresponding to a current task identifier, a primary image not associated with the current task identifier, and a primary image associated with the current task identifier. The second save button, the second previous task button, and the second next task button are displayed in a second function display area of ​​the second display device, the second jump button and the second task information are displayed in a second task display area of ​​the second display device, the primary image corresponding to the current task identifier is displayed in a third image display area of ​​the second display device, the primary image not associated with the current task identifier is displayed in a fourth image area of ​​the second display device, and the primary image associated with the current task identifier is displayed in a fifth image display area of ​​the second display device.

[0106] In the associating step, an associating process is performed on the same vehicle image displayed in the third image display area and the fourth image display area to associate the task identifiers of the same vehicle image displayed in the third image display area and the fourth image display area.

[0107] Optionally, the implementation method of associating the task identifiers of the same vehicle image displayed in the third image display area and the fourth image display area may be: by receiving an operation instruction from a user, setting the same vehicle image displayed in the fourth image display area to a third state, wherein the third state is used to determine the association relationship between the task identifier associated with the same vehicle image and the current task identifier.

[0108] Optionally, the same vehicle image displayed in the fourth image display area is set to a fourth state, and the fourth state is used to remove the association relationship between the task identifier associated with the same vehicle image and the current task identifier.

[0109] Optionally, associating the task identifiers of the same vehicle image displayed in the third image display area and the fourth image display area also includes: associating the task identifier of the vehicle image displayed in the third image display area with all task identifiers associated with the task identifier of the vehicle image displayed in the fourth image display area.

[0110] In the second confirmation step, if the current task identifier is not the last task identifier, an association step is performed based on the second functional area.

[0111] Optionally, whether the current task identifier is the last task identifier can be determined by responding to a user's operation instruction, and if the image displayed in the third image display area is not updated after receiving the user's operation instruction, the current task identifier is the last task identifier. The user's operation instruction can be a related instruction of the second next task button.

[0112] Optionally, the association relationship between the task identifiers may be saved by responding to an operation instruction of the user, wherein the operation instruction may be a related instruction of the second save button.

[0113] According to the above description, the method for merging the second task file described in this embodiment traverses the primary images in the second task file and associates the task identifiers of the same vehicle to establish an association relationship between vehicle images, thereby improving the annotation efficiency of vehicle images.

[0114] In one embodiment of the present invention, a label marking device 1000 is provided. Fig.10 The label marking device 1000 comprises:

[0115] The first vehicle image processing module 1010 is used to obtain a first vehicle image and process the first vehicle image to obtain a vehicle track and a second vehicle image, where the second vehicle image is an image bound to the vehicle track.

[0116] The first vehicle image set acquisition module 1020 is configured to classify the second vehicle images to acquire a first vehicle image set, wherein the elements in the first vehicle image set are image sets of the same vehicle.

[0117] The first task file acquisition module 1030 is used to acquire a first task file based on the first vehicle image set, wherein the first task file includes a primary image and a secondary image, and each element in the first vehicle image set has a unique task identifier.

[0118] The second task file acquisition module 1040 is used to filter the first task file to acquire a second task file.

[0119] The second task file processing module 1050 is used to merge the second task files to associate task identifiers of the same vehicle.

[0120] According to the above description, the label marking device associates the unique task identifier of the same vehicle through merging processing, thereby establishing an association relationship between the obtained new vehicle image and the old vehicle image. Therefore, the label marking device can improve the labeling efficiency of the vehicle image.

[0121] Based on the above description of the label marking method, the present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, Figure 1 The labeling method shown.

[0122] Based on the above description of the label marking method, the present invention also provides an electronic device. Fig.11 In one embodiment of the present invention, the electronic device 1100 includes: a memory 1110 on which a computer program is stored; a processor 1120, which is connected to the memory 1110 for executing the computer program and implementing Figure 1 The electronic device further comprises a display 1130, and the display 1130 is used to display a GUI interaction interface related to the labeling method.

[0123] The protection scope of the label marking method described in the present invention is not limited to the execution order of the steps listed in this embodiment. All solutions implemented by adding, reducing or replacing steps in the prior art based on the principles of the present invention are included in the protection scope of the present invention.

[0124] In summary, the labeling method, labeling device, medium and electronic device of the present invention are used to label vehicle images, which can improve the labeling efficiency of vehicle images. Therefore, the present invention effectively overcomes various shortcomings in the prior art and has high industrial utilization value.

[0125] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A label marking method, characterized in that: The label marking method comprises: Acquire a first vehicle image and process the first vehicle image to acquire a vehicle track and a second vehicle image, wherein the second vehicle image is an image bound to the vehicle track; performing classification processing on the second vehicle image to obtain a first vehicle image set, wherein the elements in the first vehicle image set are image sets of the same vehicle; Acquire a first task file based on the first vehicle image set, the first task file including primary images and secondary images, each element in the first vehicle image set having a unique task identifier, the primary image being an image with the highest average similarity to other images in the image set of the same vehicle, and the secondary image being an image remaining in the image set of the same vehicle except the primary image; Screening the first task file to obtain a second task file; The second task files are merged to associate task identifiers of the same vehicle.

2. The label marking method according to claim 1, characterized in that: Also includes: processing the second vehicle image to obtain structural similarities of adjacent images in the second vehicle image; It is determined whether the structural similarity of the adjacent images is greater than a first threshold value. If the structural similarity of the adjacent images is greater than the first threshold value, the image that is later in time sequence among the adjacent images is deleted.

3. The label marking method according to claim 1, characterized in that: The method for implementing classification processing on the second vehicle image to obtain a first vehicle image set includes: Extracting features from the second vehicle images by using a neural network model to obtain similarities between the second vehicle images; The second vehicle images are classified based on the similarities between the second vehicle images to obtain a first vehicle image set.

4. The label marking method according to claim 1, characterized in that: Also includes: Obtaining intra-class similarity of elements of the first vehicle image set; It is determined whether the intra-class similarity is greater than a second threshold or whether the intra-class similarity is less than a third threshold; if the intra-class similarity is greater than the second threshold or the intra-class similarity is less than the third threshold, filtering the elements in the first vehicle image set.

5. The label marking method according to claim 1, characterized in that: The method for obtaining the first task file based on the first vehicle image set includes: generating a unique task identifier for an element in the first vehicle image set; Obtaining pairwise similarities between images in the first vehicle image set; Processing the first vehicle image set based on pairwise similarities between the images to obtain the primary image and the secondary image, wherein the primary image is associated with a unique task identifier of an element of the first vehicle image set corresponding to the primary image, and the secondary image is associated with a unique task identifier of an element of the first vehicle image set corresponding to the secondary image; The first task file is obtained based on the primary image and the secondary image.

6. The label marking method according to claim 1, characterized in that: The method for implementing obtaining the second task file includes: a first display step: using a first display device to display a first save button, a first previous task button, a first next task button, a first jump button, first task information, and a primary image and a secondary image corresponding to a current task identifier, wherein the primary image corresponding to the current task identifier is displayed in a first image display area of ​​the first display device, the primary image corresponding to the current task identifier and the secondary image corresponding to the current task identifier are displayed in a second image display area of ​​the first display device, the first save button, the first previous task button, and the first next task button are displayed in a first function display area of ​​the first display device, and the first jump button and the first task information are displayed in a first task display area of ​​the first display device; Filtering step: filtering the images displayed in the first image display area and the second image display area according to the filtering conditions; First confirmation step: if the current task identifier is not the last task identifier, go to the filtering step, otherwise obtain the second task file.

7. The label marking method according to claim 1, characterized in that: The method for implementing the merging process on the second task file includes: A second display step: using a second display device to display a second save button, a second previous task button, a second next task button, a second jump button, second task information, a primary image corresponding to the current task identifier, a primary image not associated with the current task identifier, and a primary image associated with the current task identifier, wherein the second save button, the second previous task button, and the second next task button are displayed in a second function display area of ​​the second display device, the second jump button and the second task information are displayed in a second task display area of ​​the second display device, the primary image corresponding to the current task identifier is displayed in a third image display area of ​​the second display device, the primary image not associated with the current task identifier is displayed in a fourth image display area of ​​the second display device, and the primary image associated with the current task identifier is displayed in a fifth image display area of ​​the second display device; Associating step: performing an associative process on the same vehicle image displayed in the third image display area and the fourth image display area, so as to associate the task identifiers of the same vehicle image displayed in the third image display area and the fourth image display area; A second confirmation step: if the current task identifier is not the last task identifier, performing an association step based on the second function display area.

8. A label marking device, characterized in that: include: A first vehicle image processing module, used for acquiring a first vehicle image and processing the first vehicle image to acquire a vehicle track and a second vehicle image, wherein the second vehicle image is an image bound to the vehicle track; A first vehicle image set acquisition module, configured to classify the second vehicle image to acquire a first vehicle image set, wherein the elements in the first vehicle image set are image sets of the same vehicle; A first task file acquisition module is used to acquire a first task file based on the first vehicle image set. The first task file includes primary images and secondary images, each element in the first vehicle image set has a unique task identifier, the primary image is an image with the highest average similarity to other images in the image set of the same vehicle, and the secondary image is an image remaining in the image set of the same vehicle except the primary image; A second task file acquisition module, used for screening the first task file to acquire a second task file; The second task file processing module is used to merge the second task files to associate task identifiers of the same vehicle.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the label marking method according to any one of claims 1 to 7 is implemented.

10. An electronic device, characterized in that: The electronic device comprises: A memory storing a computer program; A processor, communicatively connected to the memory, and configured to execute the labeling method according to any one of claims 1 to 7 when calling the computer program; A display is communicatively connected to the processor and the memory, and is used to display a GUI interaction interface related to the label marking method.

Images