Portrait clustering test method, system, device and medium

Abstract

The present application provides a portrait clustering test method, system, device and medium, the method comprising: obtaining a data set, the data set being multiple portraits of the same person collected in different scenes; assigning the portraits in the data set to corresponding camera devices according to the number of camera devices, their geographical locations and preset time periods, forming an image representing the trajectory of the portrait in chronological order; clustering and archiving the images using the platform to be tested, and calculating the index of the clustering archive to test the clustering capabilities of each platform. The portrait clustering test scheme of the present application can quickly generate test data by constructing logical portrait trajectories. On the one hand, the precision, recall and splitting rate of the portrait clustering data can be calculated through the test data, and the actual performance of clustering archives provided by various manufacturers can be quickly compared; on the other hand, the test results can be obtained quickly, which improves the test efficiency.

Description

Technical Field

[0001] The present application relates to the field of image processing, and in particular to a portrait clustering test method, system, device and medium. Background Art

[0002] Image clustering, especially various image classification applications of large-scale face image (portrait) clusters (clustered into one file per person) data, is becoming increasingly valuable in the field of "active advance warning" in monitoring as face recognition and video surveillance technologies continue to advance.

[0003] However, the existing portrait clustering scheme adjusts parameters according to the characteristics of the on-site point capture to obtain the best clustering effect. In order to quickly determine the tuning parameters, a flexible automated indicator testing tool is needed. For example, the test of portrait image clustering relies on regular portrait trajectory data. When evaluating portrait image clustering, constructing a test portrait trajectory that conforms to the rules is often time-consuming and laborious, and the coverage is incomplete. Therefore, a test scheme based on portrait clustering testing is urgently needed. Summary of the invention

[0004] In view of the shortcomings of the prior art described above, the purpose of the present application is to provide a portrait clustering test method, system, device and medium, which are used to solve the problem that the portrait clustering test in the prior art cannot effectively and accurately obtain clustering indicators of various portrait images.

[0005] To achieve the above objectives and other related objectives, the present application provides a portrait clustering testing method, including:

[0006] Acquire a data set, where the data set is multiple portraits of the same person collected in different scenes;

[0007] Assigning the portraits in the data set to corresponding camera devices according to the number of camera devices, their geographical locations and a preset time period, to form an image representing the trajectory of the portraits in a time sequence;

[0008] The images are clustered and archived using the platform to be tested, and the indicators of the clustering archive are calculated to test the clustering ability of each platform, wherein the indicators include the recall rate, precision rate and split rate of various clustering algorithms.

[0009] Another object of the present application is to provide a portrait clustering testing device, comprising:

[0010] An acquisition module is used to acquire a data set, wherein the data set is a plurality of portraits of the same person collected in different scenes;

[0011] A data generation module, used to assign the portraits in the data set to corresponding camera devices according to the number of camera devices, their geographical locations and a preset time period, to form an image representing the trajectory of the portraits in a time sequence;

[0012] The clustering test module uses the platform to be tested to cluster and archive the images, and calculates the indicators of the clustering archive to test the clustering ability of each platform, wherein the indicators include the recall rate, precision rate and split rate of various clustering algorithms.

[0013] Another object of the present application is to provide an electronic device, comprising:

[0014] one or more processing devices;

[0015] A memory is used to store one or more programs; when the one or more programs are executed by the one or more processing devices, the one or more processing devices execute the portrait clustering test method.

[0016] Another object of the present application is to provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program is used to enable the computer to execute the portrait clustering test method.

[0017] As described above, the portrait clustering test method, system, device and medium of the present application have the following beneficial effects:

[0018] The portrait clustering test solution of the present application can quickly generate test data by constructing logical portrait trajectories. On the one hand, the precision, recall rate, and split rate of portrait clustering data can be calculated through the test data, and the actual performance of clustering archives provided by various manufacturers can be quickly compared; on the other hand, the test results can be obtained quickly, thereby improving the test efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 Shown is a flow chart of a portrait clustering testing method provided by this application;

[0020] Figure 2 The display is a wandering display diagram of a portrait clustering test area provided by the present application;

[0021] Figure 3 Shown is a complete flow chart of a portrait clustering test method provided by this application;

[0022] Figure 4 Shown is a flow chart of a portrait clustering testing method provided by this application;

[0023] Figure 5 Shown is a principle framework diagram of a portrait clustering test system provided by this application;

[0024] Figure 6 Shown is a structural block diagram of a portrait clustering test system provided by the present application;

[0025] Figure 7 Shown is a structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0026] The following describes the embodiments of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. The present application 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 application. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0027] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present application, and thus the drawings only show components related to the present application 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 at will, and the component layout may also be more complicated.

[0028] There are two main implementation schemes for portrait clustering (clustering archiving). The first one uses the feature vector generated by the convolutional neural network face recognition algorithm combined with the temporal and spatial features to compare similarities to obtain the clustering results of the same person (the feature vector of the same target face has high similarity). This method is called strategic clustering. The second one is GCN (Graph Convolutional Networks) - graph neural network portrait clustering (born in 2016). GCN extracts image features to generate nodes, and multiple node sets form a graph. By calculating the nodes on the classification graph (the commonly used method is Euclidean distance, and the nodes of the same person are adjacent in Euclidean distance), portrait clustering can be performed. This method is called GCN clustering. Usually, the two methods are combined to achieve the best effect. When developing clustering solutions, fast-iteration strategy solutions and GCN model changes need to be supported by automated test indicator results, otherwise it is impossible to determine which manufacturer's clustering method is better.

[0029] The testing of portrait image clustering relies on regular portrait trajectory data. When evaluating portrait image clustering, it is necessary to construct test portrait trajectories that conform to the rules, which is often time-consuming and laborious, and has incomplete coverage. In order to quickly evaluate the actual performance of portrait image clustering, it is necessary to flexibly adjust and organize the data access platform according to the parameters required by the user. Therefore, a testing scheme based on portrait clustering testing is urgently needed.

[0030] See also Figure 1, which is a flow chart of a portrait clustering test method provided in this application, including:

[0031] Step S1, obtaining a data set, wherein the data set is a plurality of portraits of the same person collected in different scenes, and the data set is obtained after being authorized and confirmed by a specific group of people;

[0032] Step S2, assigning the portraits in the data set to corresponding camera devices according to the number of camera devices, their geographical locations and a preset time period, to form an image representing the trajectory of the portraits in time sequence;

[0033] Step S3, clustering and archiving the images using the platform to be tested, and calculating the indexes of the clustering and archiving to test the clustering ability of each platform, wherein the indexes include the recall rate, precision rate and split rate of various clustering algorithms.

[0034] It should be noted that a portrait can be either a facial image or a body image. In this application, taking facial images as an example, the data set contains at least 5 facial images of the same person collected in different scenes with authorization and confirmation. The number of portraits in the data set must be greater than the number of camera devices so that the generated portrait trajectory can cover the areas where each camera device is located.

[0035] In this embodiment, test data can be quickly generated by constructing logical portrait trajectories. On the one hand, the precision, recall rate, and split rate of portrait clustering data can be calculated through the test data, and the actual performance of clustering archives provided by various manufacturers can be quickly compared; on the other hand, the test results can be obtained quickly, thereby improving the test efficiency.

[0036] See also Figure 2 , a portrait clustering test area wandering display diagram provided by this application, including:

[0037] Portrait image clustering - frequent appearance analysis example:

[0038] Under the premise of authorization by the target person, the actual situation of target person A being captured by multiple checkpoints in community A is as follows:

[0039] Checkpoint 1: 10 shots were taken between 9:00 and 13:00, and 9 shots were successfully collected.

[0040] Checkpoint 2: 13 shots were taken between 10:00 and 14:00, and 12 shots were successfully collected

[0041] Checkpoint 3: 7 shots were taken between 13:00 and 14:00, and 7 shots were successfully collected.

[0042] The page searches for wandering around the area between 9:00 and 14:00 on the same day. Based on the snapshot ID, it is found that one of the suspects wandered around the area 29 times, of which 28 were captured by A (the ID returned by the interface is the database ID corresponding to the snapshot of A in the test set), and 1 was not captured by A (does not belong to the database ID corresponding to the snapshot of A in the test set). Therefore, the accuracy of the number of times target A wandered around the area is 28 / 29=96.5%, and the recall rate is 28 / 30=93.3%. Similarly, multiple targets are tested to obtain the comprehensive accuracy and recall rate of regional wandering image clustering.

[0043] Example 1

[0044] See also Figure 3 , which is a complete flow chart of a portrait clustering test method provided in this application, including:

[0045] Read the card access device information in advance, for example, the global card access device ID, location, and quantity, and obtain the map distance between each card access device according to the card access device information;

[0046] At the same time, a test set image is read in, the global image is assigned an image ID according to a unique identifier, and the image information is encoded and preprocessed to meet the preset specifications;

[0047] According to the time interval input and scene definition, the interval time of each card port device is divided according to the scene;

[0048] Check the rationality of the allocated images, allocate the images to the camera devices according to time, aggregate the images to generate image stream push information, calculate the indicators of the image stream push information, and obtain the test results of each platform.

[0049] In this embodiment, a test set is generated so that various manufacturers can provide actual performance of cluster archiving.

[0050] Example 2

[0051] See also Figure 4 , a flow chart of a portrait clustering image clustering test method provided by the present application includes:

[0052] Before the test begins, you need to prepare N groups of test set images that meet the following conditions:

[0053] 1) A test set contains only ≥5 snapshots of the same person in different scenes;

[0054] 2) The snapshot of the same person can only appear in one test set and cannot appear in other test sets;

[0055] 3) The capture scenes should cover the capture of various situations of the target solution points as much as possible, and include as many scenes as possible such as daytime, nighttime, and rainy days;

[0056] 4) Test set snapshot standards: The average person can clearly identify them as the same person.

[0057] By collecting facial images at different times and in different scenes, the richness and completeness of the data set can be ensured, thereby improving the accuracy of subsequent facial image recognition.

[0058] Aggregate index test:

[0059] 1) Each test set is stored in a folder, and the test annotated data set is imported. The system assigns a globally unique ID (corresponding to the file name) to the image.

[0060] Among them, each of the data sets is provided with a unique first identifier; and each portrait in the data set corresponds to a unique second identifier, wherein the first identifier is associated with the second identifier. For example, the GA1400 interface can be used to associate the first identifier with the second identifier. The test system can quickly find and locate the ownership of the facial image through the identifier, and can also improve the test efficiency and accuracy.

[0061] 2) Select a batch of points, randomly select some points in each picture of the same person, generate the map push order in chronological order, and determine the path distance between two points through the map interface divided by a reasonable travel speed after the camera time interval;

[0062] Specifically, virtual monitoring points, monitoring areas and landing areas are set corresponding to the geographical locations of the checkpoint devices, and the data sets corresponding to the target objects are allocated to the checkpoint devices at the corresponding geographical locations according to the image clustering requirements to form images with spatiotemporal information.

[0063] Among them, by forming a data set with images containing spatiotemporal information, it is possible to comprehensively cover each monitoring area and truly reflect the trajectory data of the target object, thereby reducing the testing difficulty of portrait clustering and improving the operability and efficiency of the test.

[0064] Specifically, the moving speed of the target object corresponding to the portrait is calculated based on the path distance of the map data and the image clustering logic. By calculating the moving speed of the target object in different image clustering detection scenarios, the disturbed data can be effectively eliminated and the authenticity of the data can be improved. For example, if two checkpoints are 10 kilometers apart and the portraits are distributed to the checkpoints by walking, if the time difference between the distribution to the two checkpoints is less than 1 minute, it does not conform to the spatiotemporal characteristics and is inconsistent with the authenticity. Therefore, it is conducive to the reasonable distribution of the data set, thereby improving the authenticity of the data set.

[0065] Specifically, the best quality picture is selected as the cover, and the earliest time period is assigned to the picture. According to the number of occurrences of the specified card slot device, the devices are assigned according to the input number of times or randomly assigned to the card slot devices according to the temporal and spatial rationality.

[0066] 3) Perform the above operations for each set of test data, sort all images by time, use the image stream to access the platform of the manufacturer to be tested (associate the image IDs in the manufacturer database with the image IDs in the platform described in this patent one by one), and wait for the archiving process to be completed.

[0067] The process of automatically calculating indicators is as follows:

[0068] 1) Precision: For N test sets, the IDs of the snapshots that successfully clustered the i-th test set are distributed in k files. The number of snapshots in the cluster of each snapshot that belongs to the same test set as it is n k The ratio n to the total number of snapshots (size(k)) of the cluster in this archive k / size(k), divided by the number of snapshots of all clustered files in this test (Num(clustered)) and averaged.

[0069]

[0070] Among them, the precision rate reflects the accuracy of the clustering algorithms of each platform (whether the clusters are accurate and whether the clustered people are the same person).

[0071] 2) Recall rate recall: for N test sets, the IDs of the snapshots of the i-th test set that are successfully clustered are distributed in k files. The ratio n of the number of snapshots in the same test set as the snapshot in the cluster of the files where the snapshot is located to the total number of snapshots in the test set i (size(i)) is k / size(i), divided by the number of snapshots of all test sets in this test (Num(all)) to get the average.

[0072]

[0073] Among them, the recall rate reflects the excellence of the clustering algorithms of each platform.

[0074] The recall rate reflects the comprehensiveness of the clustering algorithm's clustering ability (all that should be clustered together are clustered together, and the clustered classes contain the comprehensiveness of pre-calibrated data)

[0075] 3) Split rate Expansion: For the k snapshot IDs that are successfully clustered in test set i, the number of archive clusters Expandi that belong to the same test set is divided by the number of snapshots in this test (Num(clustered)) and the average is taken.

[0076]

[0077] Among them, the splitting rate can be used to quickly calculate whether the images generated by the same target object in the data set are split into multiple data sets, and the accuracy of the generated data test set can be statistically analyzed.

[0078] In this embodiment, by automatically outputting the archive status of the snapshot distribution of all test cluster files, the correct number of snapshots in each file, and comprehensive indicators (precision, recall rate, and splitting rate of various clustering algorithms), the calibrated image data distribution is organized according to the cluster file ID distribution to generate intuitive image data clustering results, which is convenient for inspection and strategy analysis.

[0079] Example 3

[0080] like Figure 5 As shown, it is a principle framework diagram of a portrait clustering test system provided by this application, including:

[0081] Using the data from the calibration test set, we customized the generation of image stream data that meets the temporal and spatial conditions and is close to the actual situation. We connected it to each platform to generate trajectory data, and calculated the precision, recall, and splitting rate of the aggregated files. At the same time, we automated the data input and indicator result output test process.

[0082] The test data organization layer, in conjunction with the operating system, is used to organize and store test image data and test json data (including device and spatiotemporal information, as well as other parameters required for different image clustering). The annotation data can be customized to import and use customized annotation data to test image clustering indicators.

[0083] Through the built-in image clustering logic and map data path distance interface, the movement speed of personnel can be flexibly defined according to the actual situation in different regions, virtual monitoring points can be set on the map, and monitoring areas and landing areas can be defined. The system automatically uses the captured images calibrated in the data layer according to the information set by the user to generate json data containing image ID, spatiotemporal information, and image push sequence and time interval, and provides standard GA1400 image stream data access through the external interaction layer.

[0084] Through the external interactive module, the precision, recall and expansion indicators under the industry unified standard are calculated based on the generated image ID and the archive distribution of the aggregated data (see Figure 4 ). When comparing the aggregation strategies of different manufacturers, the feasibility of indicator comparison is provided according to the aggregation data organization forms of different manufacturers.

[0085] In other embodiments, the moving speed of the target object corresponding to the portrait is calculated using the path distance of the map data and the time corresponding to the portrait assigned by each camera device; and whether the moving speed of the target object corresponding to the portrait conforms to the spatiotemporal rationality is detected according to the scenes corresponding to various image clustering logics; wherein, the moving speed threshold range of the current target object is determined using the scenes corresponding to the image clustering logic, and whether the moving speed of the target object is within the moving speed threshold range is calculated to determine whether the generated trajectory data of the target object conforms to the spatiotemporal rationality.

[0086] For example, assuming that in a walking scene, the average person's walking speed is 1m / s-1.9m / s, and the distance between points AB is S ab , then a reasonable snapshot is: the algorithm determines that it is the same person, and the time interval between snapshots satisfies t>S ab / 1.9.

[0087] For example, in running scenes and riding non-motor vehicle scenes, the above method can also be used to estimate the moving speed range to eliminate unreasonable snapshots and further test the accuracy of the clustering algorithm close to the actual situation.

[0088] For example, due to the uneven terrain in a certain area, different travel route choices and other comprehensive factors, deviations in the judgment of spatial and temporal rationality can be caused, which can be calibrated through algorithms.

[0089] The map data layer encapsulates well-known third-party map data packages. The map contains many positioning points (camera devices) (latitude and longitude information (J, W), altitude information H).

[0090] Specifically, based on the longitude information, latitude information and altitude information of each checkpoint device in the map data, an ant colony algorithm with speed distribution is used to determine the path to simulate the time consumed by the target object to move in the path in real situations; the time consumed by the target object from the starting point to the end point is accumulated to obtain the time consumption distribution, and the time consumption distribution is compared with the preset time range to determine whether the target object meets the spatiotemporal rationality; the algorithm is used to evaluate the spatiotemporal rationality of multiple paths of two checkpoint devices (points), for example:

[0091] a Each personnel has an initial moving speed and an initial moving speed distribution probability function.

[0092]

[0093] Set the speed range to the possible moving speeds of the persons of interest (as wide as possible), accurate to one decimal place, calculate the probability of each speed within the range (step length 0.1m / s) according to the probability function, normalize the probability value, and multiply it by the number of persons to get the number of persons with a specific moving speed, thereby obtaining the initial moving speed of each person.

[0094] Personnel b starts from the starting point and moves along the constructed graph path, releasing pheromones on the path. The pheromone concentration is inversely proportional to the distance. If a path without pheromones is encountered, the moving direction is randomly selected (the nodes that have been visited are not taken). If a path with pheromones is encountered, the path with high pheromone concentration is taken. After a period of time, the pheromone concentration on the shortest path is the highest.

[0095] c. Every time a personnel reaches a positioning point of a checkpoint device, he / she resets the moving speed to the initial speed.

[0096] d After selecting the direction of the next point, use the ant colony algorithm to correct the speed of a single straight line in the path, correct the moving speed according to the slope, and record the required time as

[0097] In the formula, T is the time required between the bayonet devices, S ab is the path distance between the bayonet devices, V0 is the initial moving speed of the bayonet device, H a , H b is the slope height between the bayonet devices, when H a -H b When H is greater than zero, the coefficient is positive. a -H b When it is less than zero, the coefficient is negative.

[0098] e accumulates T to get the total time taken by each person from the starting point to the end point, and obtains the time consumption distribution.

[0099] fThe time range falling within the above interval is considered to be reasonable in time and space.

[0100] In this embodiment, map data is used to generate trajectory data that conforms to the rationality of real time and space, ensuring the authenticity of portrait trajectory data, which is conducive to comparing the actual performance of clustering archives provided by various manufacturers.

[0101] Example 4

[0102] See also Figure 6 , is a structural block diagram of a portrait clustering test system provided by this application; including:

[0103] An acquisition module 1 is used to acquire a data set, wherein the data set is a plurality of portraits of the same person collected in different scenes;

[0104] Data generation module 2, used to assign the portraits in the data set to corresponding camera devices according to the number of camera devices, their geographical locations and a preset time period, to form an image representing the trajectory of the portraits in time sequence;

[0105] Each of the data sets is provided with a unique first identifier; and each portrait in the data set corresponds to a unique second identifier, wherein the first identifier is associated with the second identifier.

[0106] The portrait in the portrait trajectory includes image ID, spatiotemporal information, picture stream sequence and time interval.

[0107] The clustering test module 3 uses the platform to be tested to cluster and archive the images, and calculates the index of the clustering archive to test the clustering ability of each platform.

[0108] The portrait includes a face image or a body image, and the indicators include the recall rate, precision rate and split rate of the image.

[0109] Wherein, based on the above embodiment, it also includes:

[0110] The speed calculation module 4 is used to calculate the moving speed of the target object corresponding to the portrait according to the path distance of the map data and the image clustering logic.

[0111] Here, it should be noted that the clustering test module 3 uses the geographical location of the checkpoint device to set up virtual monitoring points, monitoring areas and landing areas, and according to the image clustering requirements, the data set corresponding to the target object is allocated to the checkpoint device at the corresponding geographical location to form an image with spatiotemporal information.

[0112] Among them, it should also be noted that there is a one-to-one correspondence between the portrait clustering test system and the portrait clustering test method. Here, the technical details and technical effects involved in each module and the above-mentioned process steps are the same, and there is no need to elaborate on them one by one. Please refer to the above-mentioned portrait clustering test method.

[0113] Example 5

[0114] Reference below Figure 7 , which shows a structural schematic diagram of an electronic device (such as a terminal device or a server 700) suitable for implementing an embodiment of the present disclosure. The terminal device in the embodiment of the present disclosure may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 7 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0115] like Figure 7As shown, the electronic device 700 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage device 708 into a random access memory (RAM) 703. In the RAM 703, various programs and data required for the operation of the electronic device 700 are also stored. The processing device 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 707 is also connected to the bus 704.

[0116] Typically, the following devices may be connected to the I / O interface 707: input devices 707 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; output devices 707 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 708 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 709. The communication devices 709 may allow the electronic device 700 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 7 The electronic device 700 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.

[0117] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication device 709, or installed from the storage device 708, or installed from the ROM 702. When the computer program is executed by the processing device 701, the above functions defined in the method of the embodiment of the present disclosure are executed.

[0118] It should be noted that the computer-readable medium disclosed above may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0119] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0120] The above-mentioned computer-readable medium carries one or more programs. When the above-mentioned one or more programs are executed by the electronic device, the electronic device: obtains a data set, which is multiple portraits of the same person collected in different scenes after authorization by a specific group of people; allocates the portraits in the data set to corresponding camera devices according to the number of camera devices, their distributed geographical locations and preset time periods, forming an image that represents the trajectory of the portrait in time sequence; uses the platform to be tested to cluster and archive the images, and calculates the indicators of the cluster archiving.

[0121] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0122] The flow chart and block diagram in the accompanying drawings illustrate the possible implementation architecture, function and operation of the method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0123] To sum up, the portrait clustering test scheme of the present application can quickly generate test data by constructing logical portrait trajectories. On the one hand, the precision, recall rate, and split rate of portrait clustering data can be calculated through the test data, and the actual performance of clustering archives provided by various manufacturers can be quickly compared; on the other hand, the test results can be obtained quickly, thereby improving the test efficiency.

[0124] The above embodiments are merely illustrative of the principles and effects of the present application and are not intended to limit the present application. Anyone familiar with the technology may modify or change the above embodiments without violating the spirit and scope of the present application. Therefore, all equivalent modifications or changes made by a person of ordinary skill in the art without departing from the spirit and technical ideas disclosed in the present application shall still be covered by the claims of the present application.

Claims

1. A portrait clustering test method, characterized in that: The method comprises the following steps: Acquire a data set, where the data set is multiple portraits of the same person collected in different scenes; Assigning the portraits in the data set to corresponding camera devices according to the number of camera devices, their geographical locations and a preset time period, to form an image representing the trajectory of the portraits in a time sequence; The images are clustered and archived using the platform to be tested, and the indicators of the clustering archive are calculated to test the clustering ability of each platform, wherein the indicators include the recall rate, precision rate and split rate of various clustering algorithms.

2. The portrait clustering test method according to claim 1, characterized in that: Each of the data sets is provided with a unique first identifier; and each portrait in the data set corresponds to a unique second identifier, wherein the first identifier is associated with the second identifier.

3. The portrait clustering test method according to claim 1 or 2, characterized in that: The image of the portrait trajectory includes image ID, spatiotemporal information, picture stream sequence and time interval.

4. The portrait clustering test method according to claim 3, characterized in that: The virtual monitoring points, monitoring areas and landing areas are set according to the geographical location of the checkpoint device, and the data set corresponding to the target object is allocated to the checkpoint device at the corresponding geographical location according to image clustering to form an image with spatiotemporal information.

5. The portrait clustering test method according to claim 1, characterized in that: Also includes: The moving speed of the target object corresponding to the portrait in different scenes is calculated according to the path distance of the map data and the image clustering logic.

6. The portrait clustering test method according to claim 5, characterized in that: The step of calculating the moving speed of the target object corresponding to the portrait in different scenes according to the path distance of the map data and the image clustering logic includes: The moving speed of the target object corresponding to the portrait is calculated by using the path distance of the map data and the time corresponding to the portrait assigned by each camera device; whether the moving speed of the target object corresponding to the portrait conforms to the temporal and spatial rationality is detected according to the scenes corresponding to various image clustering logics; wherein, the moving speed threshold range of the current target object is determined by using the scenes corresponding to the image clustering logic, and whether the moving speed of the target object is within the moving speed threshold range is calculated to determine whether the generated trajectory data of the target object conforms to the temporal and spatial rationality.

7. The portrait clustering test method according to claim 1 or 6, characterized in that: Also includes: Based on the longitude information, latitude information and altitude information of each card access device in the map data, the path is determined by using the ant colony algorithm with speed distribution to simulate the time taken by the target object to move within the path in real situations; The time consumed by the target object from the starting point to the end point is accumulated to obtain a time distribution, and the time distribution is compared with a preset time range to determine whether the target object meets the time-space rationality; wherein, the speed of a single straight line in the path is corrected using an ant colony algorithm, and the ant colony algorithm is specifically as follows: In the formula, T is the time required between the bayonet devices, S ab is the path distance between the bayonet devices, V0 is the initial moving speed of the bayonet device, H a , H b is the slope height between the bayonet devices, when H a -H b When H is greater than zero, the coefficient is positive. a -H b When it is less than zero, the coefficient is negative.

8. A portrait clustering test system, characterized in that: The system comprises: An acquisition module is used to acquire a data set, wherein the data set is a plurality of portraits of the same person collected in different scenes; A data generation module, used to assign the portraits in the data set to corresponding camera devices according to the number of camera devices, their geographical locations and a preset time period, to form an image representing the trajectory of the portraits in a time sequence; The clustering test module uses the platform to be tested to cluster and archive the images, and calculates the indicators of the clustering archive to test the clustering ability of each platform, wherein the indicators include the recall rate, precision rate and split rate of various clustering algorithms.

9. An electronic device, characterized in that: include: one or more processing devices; A memory for storing one or more programs; When the one or more programs are executed by the one or more processing devices, the one or more processing devices implement the portrait clustering test method as described in any one of 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The computer program is used to enable the computer to execute the portrait clustering test method described in any one of claims 1 to 7.

Images