An image processing method and apparatus

Abstract

This invention discloses an image processing method and apparatus for making reasonable use of low-quality images to improve the completeness of archives. The method includes: dividing multiple images into low-quality and high-quality images; clustering the high-quality images based on identity information to obtain high-quality sub-archives corresponding to each identity information; clustering the low-quality images based on vehicle identifiers to obtain low-quality sub-archives corresponding to each vehicle identifier; labeling the low-quality sub-archives corresponding to vehicle identifiers with identity information; and merging low-quality sub-archives and high-quality sub-archives with the same identity information to obtain archives corresponding to the identity information. This method makes reasonable use of low-quality images to improve the completeness of archives.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to an image processing method and apparatus. Background Technology

[0002] As national security and prevention measures continue to escalate, intelligent image capture devices are covering an increasing number of scenarios. However, due to the uncontrollable nature of these scenarios, low-quality images are frequently captured. For example, weather conditions can cause variations in image brightness, leading to underexposure or overexposure; insufficient lighting at night can result in significant image noise; and when a person is moving, the captured image may be severely blurry.

[0003] In facial recognition clustering schemes, low-quality images are typically discarded to reduce data storage and computational resource consumption. However, if a file lacks data related to these low-quality images, the file is incomplete and may affect subsequent applications. For example, discarding low-quality images captured by traffic cameras may result in incomplete vehicle trajectories.

[0004] Therefore, how to make reasonable use of low-quality images to make the archives more complete is a technical problem that needs to be solved. Summary of the Invention

[0005] This invention provides an image processing method and apparatus for making reasonable use of low-quality images to make archives more complete.

[0006] To achieve the above objectives, embodiments of the present invention disclose an image processing method, the method comprising:

[0007] Acquire multiple images;

[0008] The multiple images are divided into a first quality image and a second quality image, wherein the second quality image is superior to the first quality image.

[0009] Clustering is performed on the second quality image based on identity information to obtain the second quality sub-file corresponding to each identity information;

[0010] The first quality image is clustered based on vehicle identifiers to obtain a first quality sub-file corresponding to each vehicle identifier; identity information is then labeled for the first quality sub-files corresponding to the vehicle identifiers.

[0011] The first quality sub-file and the second quality sub-file with the same identity information are merged to obtain the file corresponding to the identity information.

[0012] In one example, it further includes: obtaining relevant information about the image, wherein the relevant information about the image includes one or more of the following: checkpoint identification, time information, latitude and longitude information, scene information, coordinate information, quality score, sharpness, pitch angle, width, and confidence level.

[0013] In one example, dividing the multiple images into a first quality image and a second quality image includes: dividing the multiple images into a first quality image and a second quality image based on some relevant information of the image and a threshold corresponding to the relevant information; the relevant information includes one or more of the following: quality score, sharpness, pitch angle, width, and confidence level.

[0014] One example includes, before labeling the identity information of the first quality sub-file corresponding to the vehicle identifier, the method further includes: for any vehicle identifier, determining the first quality trajectory corresponding to the vehicle identifier based on the first quality sub-file corresponding to the vehicle identifier; comparing the first quality trajectory corresponding to the vehicle identifier with the accurate trajectory corresponding to the vehicle identifier, and determining that the similarity is greater than or equal to a set threshold. Another example involves clustering the second quality image based on identity information to obtain a second quality sub-file corresponding to each identity information, including: clustering the second quality image based on facial information to obtain a second quality facial sub-file corresponding to each facial information; clustering the second quality image based on human body information to obtain a second quality human body sub-file corresponding to each human body information; and identifying the second quality human body sub-file using facial information; merging the second quality facial sub-files and second quality human body sub-files with the same facial information to obtain the second quality sub-file corresponding to the facial information; searching for identity information associated with the facial information, and associating the identity information with the second quality sub-file corresponding to the facial information to obtain a second quality sub-file corresponding to each identity information.

[0015] In one example, before determining the first quality trajectory corresponding to the vehicle identifier based on the first quality sub-file corresponding to the vehicle identifier, the method further includes: determining overlapping vehicle identifiers that are jointly corresponding to the first quality sub-file and the second quality image, and filtering out the first quality sub-files corresponding to the overlapping vehicle identifiers from the first quality sub-file.

[0016] In one example, marking identity information for the first quality sub-file corresponding to the vehicle identifier includes: querying identity information associated with the vehicle identifier; if one identity information is found, marking the found identity information for the first quality sub-file corresponding to the vehicle identifier; if multiple identity information is found, comparing the first quality trajectory corresponding to the vehicle identifier with the second quality trajectories corresponding to the multiple identity information respectively, determining the target identity information corresponding to the trajectory with the highest similarity, marking the target identity information for the first quality sub-file corresponding to the vehicle identifier, and determining the second quality trajectory based on the second quality file.

[0017] This application provides an image processing apparatus, the apparatus comprising:

[0018] The acquisition module is used to acquire multiple images;

[0019] The processing module is used to divide the multiple images into a first quality image and a second quality image, wherein the second quality image is better than the first quality image; to cluster the second quality image based on identity information to obtain a second quality sub-file corresponding to each identity information; to cluster the first quality image based on vehicle identification to obtain a first quality sub-file corresponding to each vehicle identification; and to label the first quality sub-file corresponding to the vehicle identification with identity information.

[0020] The merging module is used to merge the first quality sub-file and the second quality sub-file that have the same identity information to obtain the file corresponding to the identity information.

[0021] In one example, the acquisition module is further configured to acquire relevant information about the image, wherein the relevant information about the image includes one or more of the following: checkpoint identification, time information, latitude and longitude information, scene information, coordinate information, sharpness, pitch angle, width, and confidence level.

[0022] In one example, when the processing module is used to divide the multiple images into a first quality image and a second quality image, it is specifically used to: divide the multiple images into a first quality image and a second quality image based on some relevant information of the image and the threshold corresponding to the relevant information; the relevant information includes one or more of the following: sharpness, pitch angle, width, and confidence.

[0023] In one example, the processing module is further configured to, for any vehicle identifier, determine a first quality trajectory corresponding to the vehicle identifier based on a first quality sub-file corresponding to the vehicle identifier; compare the first quality trajectory corresponding to the vehicle identifier with the accurate trajectory corresponding to the vehicle identifier, and determine that the similarity is greater than or equal to a set threshold. In another example, when the processing module is used to cluster the second quality image based on identity information to obtain a second quality sub-file corresponding to each identity information, it is specifically configured to: cluster the second quality image based on facial information to obtain a second quality facial sub-file corresponding to each facial information; cluster the second quality image based on human body information to obtain a second quality human body sub-file corresponding to each human body information; and identify the second quality human body sub-file using facial information; merge the second quality facial sub-files and second quality human body sub-files with the same facial information to obtain a second quality sub-file corresponding to the facial information; search for identity information associated with the facial information, and associate the identity information with the second quality sub-file corresponding to the facial information to obtain a second quality sub-file corresponding to each identity information.

[0024] In one example, it further includes: a filtering module for determining overlapping vehicle identifiers that are jointly represented by the first quality sub-file and the second quality image, and filtering out the first quality sub-file corresponding to the overlapping vehicle identifiers from the first quality sub-file.

[0025] In one example, when the processing module is used to mark identity information for the first quality sub-file corresponding to the vehicle identifier, it specifically performs the following:

[0026] The system queries the identity information associated with the vehicle identifier. If one identity information is found, it marks the found identity information in the first quality sub-file corresponding to the vehicle identifier. If multiple identity information is found, the first quality trajectory corresponding to the vehicle identifier is compared with the second quality trajectories corresponding to the multiple identity information, and the target identity information corresponding to the trajectory with the highest similarity is determined. The target identity information is then marked in the first quality sub-file corresponding to the vehicle identifier, and the second quality trajectory is determined based on the second quality file.

[0027] This application provides an image processing apparatus, including a processor and a memory;

[0028] The memory is used to store computer programs or instructions;

[0029] The processor is configured to execute some or all of the computer programs or instructions in the memory, and when the some or all of the computer programs or instructions are executed, to implement the image processing method described above.

[0030] This application provides a computer-readable storage medium for storing a computer program, the computer program including instructions for implementing the above-described method.

[0031] This application provides a computer program product comprising: computer program code, which, when run on a computer, causes the computer to perform the above-described method.

[0032] Although the low-quality images in this application are not included in facial image clustering (e.g., clustering based on identity information), they are not discarded either. Instead, the low-quality images are clustered by vehicle identifiers to obtain low-quality archives, which are then identified with identity information. This allows the low-quality archives obtained through vehicle identifier clustering to be associated with the high-quality archives obtained through facial image clustering using the identity information. The entire process ensures that images discarded in facial image clustering (i.e., low-quality images) are properly archived, and because they are not included in facial image clustering, they do not affect the centroid of the archives (the reference or representative images of the archives). Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 A schematic diagram of an image processing procedure provided in this application;

[0035] Figure 2 A schematic diagram of an image processing procedure provided in this application;

[0036] Figure 3 A schematic diagram of an image processing procedure provided in this application;

[0037] Figure 4 A schematic diagram of an image processing procedure provided in this application;

[0038] Figure 5 A structural diagram of an image processing device provided in this application;

[0039] Figure 6 This application provides a structural diagram of an image processing device. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] Figure 1 This application provides a schematic diagram of an image process, the execution device of which can be any electronic device. The process includes the following steps:

[0042] Step 11: Acquire multiple images.

[0043] This application primarily focuses on images of people driving on roads. The acquired images are used to represent people driving on roads, and these images can also be referred to as vehicle window images. Multiple images may contain, but are not limited to, one or more of the following information: faces, human bodies, and vehicle identification. Vehicle identification includes, for example, license plate numbers, vehicle colors, and vehicle brands. This application allows for the deployment of capture or recording devices at various traffic intersections, checkpoints, roads, schools, hospitals, and other locations to acquire multiple images or select multiple frames from a video.

[0044] Optionally, relevant image information can also be obtained. This information includes one or more of the following: checkpoint identification, acquisition time, latitude and longitude information, scene information, coordinate information (e.g., coordinates of a face / body / license plate in the image), quality score (e.g., weighted value for confidence), image sharpness, image pitch angle, image width, image confidence, and the identification of the device that acquired the image. Image sharpness, image pitch angle (the angle relative to a person as they move, the orientation of the face, as determined by image parsing), image width, and image coordinates can be considered image attribute information. Checkpoint identification, spatiotemporal information (e.g., time, latitude and longitude), and scene information (e.g., school, village, hospital, bank) can be used to determine the trajectory. As a person or vehicle moves (e.g., turning, looking back, looking to the side), the orientation of the face, body, or license plate changes; this pitch angle can refer to the orientation of the face, body, or license plate.

[0045] The process of step 11 will be explained in detail later.

[0046] Step 12: Divide the multiple images into a first (e.g., low) quality image and a second (e.g., high) quality image.

[0047] It should be noted that the first quality and second quality mentioned in this application are used for comparison of image quality, with the second quality being superior to the first quality. "First" and "second" can also be replaced with other terms that can represent image quality, such as "first" being replaced with "low" or "low-medium," and "second" being replaced with "high" or "high-medium." For ease of understanding, the following explanation will use the first quality as low quality and the second quality as high quality as an example.

[0048] When distinguishing images of different qualities, multiple images can be categorized into low-quality and high-quality images based on partially relevant information of the images and the corresponding thresholds. The partially relevant information includes one or more of the following: quality score, sharpness, pitch angle, image width, and confidence level.

[0049] Step 13: Cluster the high-quality images based on the identity information to obtain high-quality sub-files corresponding to each identity information.

[0050] A high-quality sub-file corresponding to any identity information includes a high-quality image corresponding to the identity information and related information about the high-quality image.

[0051] Identity information can be facial information or ID card information.

[0052] High-quality images may include facial images or human images. The following describes the process of clustering these high-quality images based on identity information to obtain high-quality sub-files corresponding to each identity information:

[0053] Clustering is performed on the high-quality images based on facial information to obtain high-quality facial sub-files corresponding to each facial information;

[0054] The high-quality images are clustered based on human body information to obtain high-quality human body sub-files corresponding to each human body information; and the high-quality human body sub-files are identified using facial information.

[0055] The high-quality face sub-file and the high-quality body sub-file with the same face information are merged to obtain the high-quality sub-file corresponding to the face information;

[0056] Find the identity information associated with the facial information, and associate the identity information with the high-quality sub-file corresponding to the facial information to obtain the high-quality sub-file corresponding to each identity information.

[0057] Additionally, it should be noted that the file includes images and related information.

[0058] Step 14: Cluster the low-quality images based on vehicle identifiers to obtain low-quality sub-files corresponding to each vehicle identifier.

[0059] The low-quality sub-file corresponding to any vehicle identifier includes the low-quality image corresponding to the vehicle identifier and related information about the low-quality image.

[0060] The order of steps 13 and 14 is not limited.

[0061] Step 15: Mark the identity information for the low-quality sub-file corresponding to the vehicle identifier.

[0062] For example, low-quality images that do not meet the threshold can be grouped by vehicle identifier to form a low-quality profile. Information such as the number of checkpoints encountered, the frequency of occurrence within a time period, and vehicle trajectory can also be statistically analyzed within the low-quality profile to obtain the low-quality trajectory corresponding to that low-quality profile (i.e., that vehicle identifier).

[0063] When marking identity information for the low-quality sub-file corresponding to the vehicle identifier, one can first search for identity information associated with the vehicle identifier and then mark the found identity information onto the low-quality sub-file corresponding to the vehicle identifier.

[0064] For example, search for the identity information associated with the vehicle identification in the static data corresponding to the vehicle identification.

[0065] If an identity information is found, it is an identity information found by marking the low-quality sub-file corresponding to the vehicle identifier.

[0066] The multiple images obtained in this application were acquired over a period of time. During this period, the vehicle may have been bought and sold, and the vehicle owner may have changed, which may result in the discovery of multiple identity information.

[0067] In one example, if multiple identity information is found, one identity information can be arbitrarily selected as the target information, or the identity information with a later time can be selected as the target information, and the target identity information can be marked in the low-quality sub-file corresponding to the vehicle identifier.

[0068] In another example, if multiple identity information is found, the low-quality trajectory corresponding to the vehicle identifier is compared with the high-quality trajectories corresponding to the high-quality files of each of the multiple identity information. The target identity information corresponding to the trajectory with the highest similarity is determined, and the target identity information is labeled in the low-quality sub-file corresponding to the vehicle identifier. Determining identity information through similarity comparison can improve the accuracy of identity determination.

[0069] Step 16: Merge the low-quality sub-files and high-quality sub-files with the same identity information to obtain the file corresponding to the identity information.

[0070] Although the low-quality images in this application are not included in facial image clustering (e.g., clustering based on identity information), they are not discarded either. Instead, the low-quality images are clustered by vehicle identifiers to obtain low-quality archives, which are then identified with identity information. This allows the low-quality archives obtained through vehicle identifier clustering to be associated with the high-quality archives obtained through facial image clustering using the identity information. The entire process ensures that images discarded in facial image clustering (i.e., low-quality images) are properly archived, and because they are not included in facial image clustering, they do not affect the centroid of the archives (the reference or representative images of the archives).

[0071] In one optional example, for any vehicle identifier, a low-quality trajectory corresponding to the vehicle identifier is determined based on the low-quality sub-file corresponding to the vehicle identifier. The low-quality trajectory corresponding to the vehicle identifier is compared with the accurate trajectory corresponding to the vehicle identifier. If the similarity is greater than or equal to a set threshold, the low-quality sub-file corresponding to the vehicle identifier is labeled with identity information.

[0072] The accurate trajectory corresponding to the vehicle identification mark can be pre-existing in static data. For example, a dedicated image capture or video recording device can be deployed to determine the vehicle's trajectory. The images used to determine the vehicle's trajectory are clear. Based on these images, the accurate trajectory corresponding to the vehicle identification mark is determined and stored in the static data corresponding to that vehicle identification mark. The device dedicated to determining the vehicle's trajectory is different from the image acquisition device in step 101 of this application, and the source of the images used to determine the vehicle's trajectory is different from the source of the images acquired in step 101 of this application.

[0073] Alternatively, this application can cluster the high-quality images in step 102 based on vehicle identifiers to obtain high-quality sub-files corresponding to each vehicle identifier; any high-quality sub-file corresponding to a vehicle identifier includes the high-quality image corresponding to the vehicle identifier and related information of the high-quality image; then, based on the high-quality sub-files corresponding to the vehicle identifiers, the high-quality trajectory corresponding to the vehicle identifiers is determined, and the high-quality trajectory is determined as the accurate trajectory.

[0074] By comparing the low-quality trajectory corresponding to a low-quality file with the accurate trajectory corresponding to a vehicle identifier, and retaining the low-quality file if the similarity meets the requirements, the low-quality file can be identified and its identity information can be added, thereby improving the accuracy of the low-quality file.

[0075] In one alternative example, before determining the low-quality trajectory corresponding to the vehicle identifier based on the low-quality sub-file corresponding to the vehicle identifier, overlapping vehicle identifiers corresponding to both the low-quality sub-file and the high-quality image can be determined, and the low-quality sub-files corresponding to the overlapping vehicle identifiers can be filtered out from the low-quality sub-files.

[0076] Alternatively, before comparing the low-quality trajectory corresponding to the vehicle identifier with the accurate trajectory corresponding to the vehicle identifier, overlapping vehicle identifiers corresponding to both low-quality sub-files and high-quality images can be identified, and the low-quality trajectories corresponding to the overlapping vehicle identifiers can be filtered out from the low-quality trajectories.

[0077] For example, this application clusters four low-quality sub-files corresponding to vehicle identifiers, such as vehicle identifier 1, vehicle identifier 2, vehicle identifier 3, and vehicle identifier 4. Four vehicle identifiers are found in the high-quality image: vehicle identifier 3, vehicle identifier 4, vehicle identifier 5, and vehicle identifier 6. Since vehicle identifiers 3 and 4 are included in both the high-quality and low-quality images, and are overlapping vehicle identifiers, the low-quality files (or low-quality trajectories) corresponding to vehicle identifiers 3 and 4 can be filtered out and no longer participate in the subsequent trajectory comparison process.

[0078] In addition, when determining the vehicle identifiers corresponding to high-quality images, the high-quality images can be clustered based on the vehicle identifiers to obtain high-quality sub-files corresponding to each vehicle identifier; thus obtaining the vehicle identifiers corresponding to the high-quality images.

[0079] like Figure 2 As shown, the relevant process of step 11: acquiring multiple images is introduced.

[0080] Step 201: Data collected by various capture devices (such as images, videos, audio, etc.) is transmitted to the front-end storage via the network through the front-end sensing device. The front-end storage is used for data.

[0081] Step 202 involves classifying the data collected by the capture device, such as into images, videos, and audio. Images captured by the capture device can be used as the images to be used in this application, and videos can also be parsed to generate the images to be used in this application. For example, videos can be optimized by selecting several frames from a video segment as the images to be used in this application. For instance, for data collected by a device that only captures faces, images containing faces can be selected; for data collected by a structured capture device, images containing faces and images containing human bodies can be selected. The selected images may also include license plates.

[0082] Understandably, the data may also contain spatiotemporal information, scene information, etc.

[0083] Step 203: The selected image is passed to the parsing operator via a message queue, such as the face operator or the body operator. After parsing, different operators will generate corresponding attribute information (e.g., sharpness, coordinates, etc.). This attribute information is then added to the message body corresponding to the image.

[0084] Attribute information, spatiotemporal information, scene information, etc. are all related information about images.

[0085] Optionally, in step 204, the image data (which can be understood as the image's feature vector (a feature vector is used to represent an image; for ease of description, this application abbreviates image feature vector as "image") and related image information) can be sent to different message queues, such as the message queues corresponding to the face stream, body stream, and vehicle stream. Alternatively, the face stream, body stream, and vehicle stream can be ignored and sent to the same message queue. It is understood that images in the face stream, body stream, and vehicle stream may overlap. For example, if an image includes both a face and a body, the image data can be sent to either the message queue corresponding to the face stream or the message queue corresponding to the body stream. Similarly, if an image includes both a face and a vehicle identifier (e.g., license plate number), the image data can be sent to either the message queue corresponding to the face stream or the message queue corresponding to the vehicle stream.

[0086] The following describes the process of dividing the multiple images into a first (low) quality image and a second (high) quality image in step 12.

[0087] Images are evaluated based on thresholds corresponding to different relevant information (e.g., quality score, sharpness, pitch angle, image width, confidence level, etc.) to distinguish between low-quality and high-quality images. Low-quality and high-quality images can be sent to different message queues.

[0088] Optionally, different quality levels can be distinguished for different types of data streams (e.g., face streams, body streams, vehicle streams), and for any given category, low-quality and high-quality images can be sent to different message queues, resulting in 6 message queues. Alternatively, the data stream category can be ignored, resulting in 2 message queues.

[0089] like Figure 3 As shown, the process of step 13, clustering the high-quality images based on identity information to obtain high-quality sub-files corresponding to each identity information, is introduced.

[0090] Step 301: Obtain high-quality face data streams, and perform face clustering on the consumed data at certain time periods to generate high-quality face sub-profiles. For example, read high-quality face data streams from the message queue at certain time periods (N (N is greater than or equal to 0) hours; generally, the longer the face clustering time, the better the effect), and use a clustering algorithm to generate high-quality face sub-profiles.

[0091] Step 302: Obtain high-quality human body data streams and perform human body clustering on the consumed data at certain time periods to generate high-quality human body sub-profiles. For example, read high-quality human body data streams from the message queue at certain time periods (N (N is greater than or equal to 0) hours; generally, the longer the human body clustering time, the better the effect), and use a clustering algorithm to generate high-quality human body sub-profile data. If the high-quality human body sub-profiles contain facial information, then the unique identifier of the high-quality human body sub-profiles is represented by a facial identifier ID.

[0092] Step 303: Utilize the relationship between faces and bodies to merge high-quality face sub-files and high-quality body sub-files. For example, from video streams to image streams, and then to different parsing operators, related face and body data can be generated in this chain (e.g., in the same scene image, the parsed body image may contain faces). By comparing the unique identifiers of the files, the related face and body files are merged to obtain a high-quality sub-file. The identity information corresponding to the face can also be retrieved, allowing for real-name authentication of the high-quality sub-file.

[0093] Optionally, high-quality subfiles containing vehicle identifiers (such as license plate numbers) can be identified, and a high-quality trajectory of the high-quality subfile can be generated.

[0094] like Figure 4 As shown, the relevant processes of steps 15 and 16 are described in detail.

[0095] Step 401: Obtain low-quality sub-files and high-quality sub-files containing license plate information.

[0096] Step 402: Retrieve data from low-quality sub-files that does not exist in high-quality sub-files through a join query. For example, retrieve data from low-quality sub-files that does not exist in medium-to-high-quality sub-files through an SQL join query. This involves identifying overlapping license plate numbers corresponding to both low-quality and high-quality sub-files, as described earlier, and then filtering out the low-quality sub-files corresponding to these overlapping license plate numbers.

[0097] Step 403: Utilize the vehicle's static data by querying the license plate number to obtain a low-quality file containing static data. For example, the vehicle's static data includes the vehicle's identity information and trajectory data. By querying the license plate number of the low-quality sub-file in relation to the vehicle's corresponding static data, the identity information corresponding to the license plate number is obtained.

[0098] If only one identity information is retrieved, the similarity between the low-quality trajectory of the low-quality sub-file and the accurate trajectory of the vehicle is compared. If the similarity threshold is met, the low-quality sub-file is labeled with the corresponding identity; otherwise, the process ends.

[0099] If multiple identity information is found, the similarity between the low-quality trajectory of the low-quality sub-file and the accurate trajectory of the vehicle is compared. If the similarity threshold is met, the high-quality trajectory corresponding to the high-quality sub-file corresponding to each identity information is found. By comparing the similarity between the low-quality trajectory and the high-quality trajectory, the identity with the highest similarity among the multiple identity information is selected and the low-quality sub-file is labeled with the identity with the highest similarity. If the threshold is not met, the process ends.

[0100] Step 404: Merge the low-quality sub-files that have already been verified with the high-quality sub-files that have already been verified. For example, merge low-quality sub-files and high-quality sub-files with the same identity. This only merges the sub-file data of the archives and does not affect the clustering module algorithm, thus avoiding the contamination of the archive centroids by low-quality images and affecting subsequent clustering.

[0101] This method utilizes license plate numbers found in low-quality facial / human images to cluster these images by license plate number. The resulting low-quality sub-files are then compared with static vehicle data (e.g., accurate trajectories), and associated with vehicle identity information. Finally, these sub-files are merged with high-quality sub-files formed from facial image clustering, grouping them into the optimal file. Simultaneously, the clustering algorithm avoids using low-quality images, preventing centroid contamination of the files. This approach effectively reduces the number of lost file trajectories, improves the clustering rate, and significantly enhances the effectiveness of subsequent intelligent applications.

[0102] like Figure 5 As shown, an image processing apparatus is provided, the apparatus comprising:

[0103] The acquisition module 501 is used to acquire multiple images;

[0104] Processing module 502 is used to divide the multiple images into first quality images and second quality images, wherein the second quality image is better than the first quality image; cluster the second quality images based on identity information to obtain a second quality sub-file corresponding to each identity information; cluster the first quality images based on vehicle identification to obtain a first quality sub-file corresponding to each vehicle identification; and label the first quality sub-file corresponding to the vehicle identification with identity information.

[0105] The merging module 503 is used to merge the first quality sub-file and the second quality sub-file with the same identity information to obtain the file corresponding to the identity information.

[0106] In one example, the acquisition module is further configured to acquire relevant information about the image, wherein the relevant information about the image includes one or more of the following: checkpoint identification, time information, latitude and longitude information, scene information, coordinate information, sharpness, pitch angle, width, and confidence level.

[0107] In one example, when the processing module 502 is used to divide the multiple images into a first quality image and a second quality image, it is specifically used to: divide the multiple images into a first quality image and a second quality image based on some relevant information of the image and the threshold corresponding to the relevant information; the relevant information includes one or more of the following: sharpness, pitch angle, width, and confidence.

[0108] In one example, the processing module 502 is further configured to, for any vehicle identifier, determine a first quality trajectory corresponding to the vehicle identifier based on a first quality sub-file corresponding to the vehicle identifier; compare the first quality trajectory corresponding to the vehicle identifier with the accurate trajectory corresponding to the vehicle identifier, and determine that the similarity is greater than or equal to a set threshold. In another example, when the processing module 502 is used to cluster the second quality image based on identity information to obtain a second quality sub-file corresponding to each identity information, it specifically performs the following: clustering the second quality image based on facial information to obtain a second quality facial sub-file corresponding to each facial information; clustering the second quality image based on human body information to obtain a second quality human body sub-file corresponding to each human body information; and identifying the second quality human body sub-file using facial information; merging the second quality facial sub-files and second quality human body sub-files with the same facial information to obtain a second quality sub-file corresponding to the facial information; searching for identity information associated with the facial information, and associating the identity information with the second quality sub-file corresponding to the facial information to obtain a second quality sub-file corresponding to each identity information.

[0109] In one example, it further includes: a filtering module 504, used to determine overlapping vehicle identifiers that are jointly corresponding to the first quality sub-file and the second quality image, and to filter out the first quality sub-file corresponding to the overlapping vehicle identifiers in the first quality sub-file.

[0110] In one example, when the processing module 502 is used to mark identity information for the first quality sub-file corresponding to the vehicle identifier, it is specifically used to:

[0111] The system queries the identity information associated with the vehicle identifier. If one identity information is found, it marks the found identity information in the first quality sub-file corresponding to the vehicle identifier. If multiple identity information is found, the first quality trajectory corresponding to the vehicle identifier is compared with the second quality trajectories corresponding to the multiple identity information, and the target identity information corresponding to the trajectory with the highest similarity is determined. The target identity information is then marked in the first quality sub-file corresponding to the vehicle identifier, and the second quality trajectory is determined based on the second quality file.

[0112] like Figure 6As shown, this application provides an image processing apparatus, including a processor 601 and a memory 602;

[0113] The memory 602 is used to store computer programs or instructions;

[0114] The processor 601 is configured to execute some or all of the computer programs or instructions in the memory, and when the some or all of the computer programs or instructions are executed, to implement the image processing method described above.

[0115] This application also provides a computer-readable storage medium storing a computer program that, when executed by a computer, causes the computer to perform the image processing method described above. Alternatively, the computer program includes instructions for implementing the image processing method described above.

[0116] This application also provides a computer program product, including: computer program code, which, when run on a computer, enables the computer to execute the image processing method provided above.

[0117] Furthermore, the processor mentioned in the embodiments of this application can be a central processing unit (CPU), a baseband processor, and the baseband processor and CPU can be integrated together or separate. It can also be a network processor (NP) or a combination of CPU and NP. The processor may further include hardware chips or other general-purpose processors. The aforementioned hardware chip can be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The aforementioned PLD can be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), and other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., or any combination thereof. The general-purpose processor can be a microprocessor or any conventional processor.

[0118] The memory mentioned in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced Synchronous DRAM (ESDRAM), Synchlink DRAM (SLDRAM), and Direct Rambus RAM (DR RAM). It should be noted that the memories described in this application are intended to include, but are not limited to, these and any other suitable types of memory.

[0119] The transceiver mentioned in the embodiments of this application may include a separate transmitter and / or a separate receiver, or the transmitter and receiver may be integrated into one unit. The transceiver can operate under the instruction of a corresponding processor. Optionally, the transmitter may correspond to a transmitter in a physical device, and the receiver may correspond to a receiver in a physical device.

[0120] Those skilled in the art will recognize that the method steps and units described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the steps and components of each embodiment have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0121] In the embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the couplings or direct couplings or communication connections shown or discussed may be indirect couplings or communication connections through some interfaces, apparatuses, or units, or they may be electrical, mechanical, or other forms of connection.

[0122] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of this application, depending on actual needs.

[0123] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0124] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0125] In this application, "and / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "Multiple" in this application refers to two or more. Furthermore, it should be understood that in the description of this application, terms such as "first" and "second" are used only for descriptive purposes and should not be construed as indicating or implying relative importance or order.

[0126] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0127] Obviously, those skilled in the art can make various modifications and variations to the embodiments of this application without departing from the spirit and scope of the embodiments of this application. Therefore, if these modifications and variations to the embodiments of this application fall within the scope of the claims of this application and their equivalents, this application also intends to include these modifications and variations.

Claims

1. An image processing method, characterized in that, The method includes: Acquire multiple images; The multiple images are divided into a first quality image and a second quality image, wherein the second quality image is superior to the first quality image; Clustering is performed on the second quality image based on identity information to obtain the second quality sub-file corresponding to each identity information; The first quality image is clustered based on vehicle identifiers to obtain a first quality sub-file corresponding to each vehicle identifier; identity information is then labeled for the first quality sub-files corresponding to the vehicle identifiers. The first quality sub-file and the second quality sub-file with the same identity information are merged to obtain the file corresponding to the identity information; Before marking the identity information for the first quality sub-file corresponding to the vehicle identifier, the method further includes: For any vehicle identifier, a first mass trajectory corresponding to the vehicle identifier is determined based on the first mass sub-file corresponding to the vehicle identifier; the first mass trajectory corresponding to the vehicle identifier is compared with the accurate trajectory corresponding to the vehicle identifier, and the similarity is determined to be greater than or equal to a set threshold.

2. The method as described in claim 1, characterized in that, The multiple images are divided into a first quality image and a second quality image, including: Based on partial relevant information of the image and the threshold corresponding to the partial relevant information, the multiple images are divided into a first quality image and a second quality image; the partial relevant information includes one or more of the following: sharpness, pitch angle, width, and confidence.

3. The method as described in claim 1, characterized in that, Clustering is performed on the second quality images based on identity information to obtain second quality sub-files corresponding to each identity information, including: Clustering is performed on the second quality image based on facial information to obtain the second quality facial sub-file corresponding to each facial information; The second-quality images are clustered based on human body information to obtain second-quality human body sub-files corresponding to each human body information; and the second-quality human body sub-files are identified using facial information. The second quality face sub-file and the second quality body sub-file with the same face information are merged to obtain the second quality sub-file corresponding to the face information; Find the identity information associated with the facial information, and associate the identity information with the second quality sub-file corresponding to the facial information to obtain the second quality sub-file corresponding to each identity information.

4. The method as described in claim 1, characterized in that, Before determining the first mass trajectory corresponding to the vehicle identifier based on the first mass sub-file corresponding to the vehicle identifier, the process further includes: Identify overlapping vehicle identifiers that correspond to both the first quality sub-file and the second quality image, and filter out the first quality sub-file corresponding to the overlapping vehicle identifiers from the first quality sub-file.

5. The method as described in claim 1, characterized in that, The first quality sub-file corresponding to the vehicle identifier is marked with identity information, including: Query the identity information associated with the vehicle identifier; If an identity information is found, it is an identity information found by marking the first quality sub-file corresponding to the vehicle identifier; If multiple identity information is found, the first quality trajectory corresponding to the vehicle identifier is compared with the second quality trajectory corresponding to each of the multiple identity information. The target identity information corresponding to the trajectory with the highest similarity is determined, and the target identity information is marked in the first quality sub-file corresponding to the vehicle identifier. The second quality trajectory is determined based on the second quality file.

6. An image processing apparatus, characterized in that, The device includes: The acquisition module is used to acquire multiple images; The processing module is used to divide the multiple images into a first quality image and a second quality image, wherein the second quality image is better than the first quality image; to cluster the second quality image based on identity information to obtain a second quality sub-file corresponding to each identity information; to cluster the first quality image based on vehicle identification to obtain a first quality sub-file corresponding to each vehicle identification; and to label the first quality sub-file corresponding to the vehicle identification with identity information. The merging module is used to merge the first quality sub-file and the second quality sub-file that have the same identity information to obtain the file corresponding to the identity information; The processing module is further configured to, for any vehicle identifier, determine the first quality trajectory corresponding to the vehicle identifier based on the first quality sub-file corresponding to the vehicle identifier; compare the first quality trajectory corresponding to the vehicle identifier with the accurate trajectory corresponding to the vehicle identifier, and determine that the similarity is greater than or equal to a set threshold.

7. The apparatus as claimed in claim 6, characterized in that, When the processing module is used to divide the multiple images into a first quality image and a second quality image, it is specifically used for: Based on partially relevant information of the images and the threshold corresponding to the partially relevant information, the multiple images are divided into a first quality image and a second quality image; the partially relevant information includes one or more of the following: quality score, sharpness, pitch angle, width, and confidence level.

8. The apparatus as claimed in claim 6, characterized in that, When the processing module clusters the second quality image based on identity information to obtain the second quality sub-file corresponding to each identity information, it is specifically used for: Clustering is performed on the second quality image based on facial information to obtain a second quality facial sub-file corresponding to each facial information; clustering is performed on the second quality image based on human body information to obtain a second quality human body sub-file corresponding to each human body information; The second quality human body sub-file is identified by facial information; the second quality facial sub-file and the second quality human body sub-file with the same facial information are merged to obtain the second quality sub-file corresponding to the facial information; Find the identity information associated with the facial information, and associate the identity information with the second quality sub-file corresponding to the facial information to obtain the second quality sub-file corresponding to each identity information.

9. The apparatus as claimed in claim 6, characterized in that, Also includes: The filtering module is used to determine the overlapping vehicle identifiers that are jointly corresponding to the first quality sub-file and the second quality image, and to filter out the first quality sub-files corresponding to the overlapping vehicle identifiers.

10. The apparatus as claimed in claim 6, characterized in that, When the processing module is used to mark identity information for the first quality sub-file corresponding to the vehicle identifier, it is specifically used for: Query the identity information associated with the vehicle identifier; if an identity information is found, it is an identity information found by the first quality sub-file tag corresponding to the vehicle identifier; If multiple identity information is found, the first quality trajectory corresponding to the vehicle identifier is compared with the second quality trajectory corresponding to each of the multiple identity information. The target identity information corresponding to the trajectory with the highest similarity is determined, and the target identity information is marked in the first quality sub-file corresponding to the vehicle identifier. The second quality trajectory is determined based on the second quality file.

11. An image processing apparatus, characterized in that, Including processor and memory; The memory is used to store computer programs or instructions; The processor is configured to execute some or all of the computer programs or instructions in the memory, and when the some or all of the computer programs or instructions are executed, to implement the method as described in any one of claims 1-5.

12. A computer-readable storage medium, characterized in that, Used to store a computer program, the computer program including instructions for implementing the method of any one of claims 1-5.

13. A computer program product, characterized in that, The computer program product includes: computer program code, which, when run on a computer, causes the computer to perform the method as described in any one of claims 1-5.

Images