Archive anomaly detection method and device, processing equipment, storage medium and chip

Abstract

The application relates to an archive anomaly detection method and device, a processing device, a storage medium and a chip. The method comprises the following steps: acquiring at least one feature image of a portrait archive, wherein the feature image comprises a human face image and / or a human body image with a similarity between the human face image and / or the human body image and a centroid of the portrait archive greater than a preset similarity threshold; determining a first similarity according to the similarity between the at least one feature image and the centroid, and / or determining a second similarity according to the similarity between the at least one feature image; and determining an anomaly probability of the portrait archive according to a time interval from a previous archive anomaly detection, the first similarity and / or the second similarity. The archive anomaly detection method provided by the application can determine whether the portrait archive is abnormal without extracting all information in the portrait archive, thereby improving the archive anomaly detection efficiency and saving a large amount of detection time.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method, apparatus, processing device, storage medium, and chip for detecting file anomalies. Background Technology

[0002] With the development of mobile internet technology and the continuous innovation of facial recognition technology, various monitoring devices have been widely used in the security field, forming facial recognition surveillance systems. These facial recognition-based surveillance systems can cluster images to create facial profiles.

[0003] In actual image clustering processes, due to various factors such as inclement weather, obstruction by foreign objects, equipment damage, and algorithm defects, the final image profiles may contain misclassified images, such as Zhang San's image profile containing Li Si's image. However, current technologies lack mechanisms to detect misclassified image profiles.

[0004] Therefore, there is an urgent need for a highly efficient method for detecting archival anomalies. Summary of the Invention

[0005] Based on this, and in response to the aforementioned technical problems, this application provides a method, apparatus, processing device, storage medium, and chip for detecting archival anomalies, in order to at least solve the problem of the lack of a detection mechanism for misplaced portrait archives in related technologies, thereby improving the efficiency of archival anomaly detection.

[0006] In a first aspect, embodiments of this application provide a method for detecting file anomalies, the method comprising:

[0007] Acquire at least one feature image of the portrait profile, the feature image including a face image and / or a human body image whose similarity to the centroid of the portrait profile is greater than a preset similarity threshold;

[0008] A first similarity is determined based on the similarity between the at least one feature image and the centroid, and / or a second similarity is determined based on the similarity between the at least one feature image and the centroid.

[0009] The probability of an anomaly in the portrait file is determined based on the time interval since the last file anomaly detection, the first similarity, and / or the second similarity.

[0010] This application provides a method for detecting anomalies in portrait archives. During anomaly detection, the similarity between the centroid of the portrait archive and a feature image, along with the time interval since the last anomaly detection, can determine the probability of an anomaly in the portrait archive. Since the feature image can be a face image and / or a human image with a similarity greater than a preset similarity threshold to the centroid of the portrait archive, it can represent the most abundant feature information in the portrait archive. This provides a more accurate benchmark for subsequent similarity calculations, thereby improving the accuracy of anomaly detection. Furthermore, this method eliminates the need to extract all information from the portrait archive to determine its anomaly status, thus improving the efficiency of anomaly detection and saving significant detection time.

[0011] Optionally, in one embodiment of this application, determining a first similarity based on the similarity between the at least one feature image and the centroid, and / or determining a second similarity based on the similarity between the at least one feature image and the centroid, includes:

[0012] Multiple first candidate similarities between the at least one feature image and the centroid are determined, and the first candidate similarity that meets the first preset requirement is determined as the first similarity; the first preset requirement includes the smallest similarity among the multiple first candidate similarities;

[0013] And / or, determine the second candidate similarity between the at least one feature image, and determine the second candidate similarity less than the second preset threshold as the second similarity; the second preset requirement includes the smallest similarity among the plurality of second candidate similarities.

[0014] Optionally, in one embodiment of this application, determining the anomaly probability of the portrait file based on the time interval since the last file anomaly detection, the first similarity, and / or the second similarity includes:

[0015] The probability of an anomaly in the portrait file is determined based on the positive correlation between the time interval since the last anomaly detection and the probability of anomaly in the portrait file, and the negative correlation between the first similarity and / or the second similarity and the probability of anomaly in the portrait file.

[0016] Optionally, in one embodiment of this application, determining the anomaly probability of the portrait file based on the positive correlation between the time interval since the last anomaly detection and the anomaly probability of the portrait file, and the negative correlation between the first similarity and / or the second similarity and the anomaly probability of the portrait file, includes:

[0017] Determine the weighted value of the first similarity and the second similarity;

[0018] The probability of an anomaly in the portrait file is determined based on the ratio of the weighted value to the time interval since the last anomaly detection.

[0019] Optionally, in one embodiment of this application, after determining the probability of anomaly in the portrait file based on the time interval since the last file anomaly detection, the first similarity, and / or the second similarity, the method further includes:

[0020] The portrait files with an anomaly probability greater than a preset anomaly probability threshold are re-archived.

[0021] Optionally, in one embodiment of this application, the step of re-archiving the portrait files whose anomaly probability is greater than a preset anomaly probability threshold includes:

[0022] Identify at least one target portrait file whose anomaly probability is greater than a preset anomaly probability threshold;

[0023] The confidence level of each of the at least one target portrait file is determined, wherein the confidence level is used to characterize the degree of correlation between the human body image in the target portrait file and the face image in the portrait file;

[0024] Target portrait files with a confidence level lower than a preset confidence threshold are preferentially re-archived.

[0025] Optionally, in one embodiment of this application, determining the anomaly probability of the portrait file based on the time interval since the last file anomaly detection, the first similarity, and / or the second similarity includes:

[0026] Obtain the historical anomaly probability of the aforementioned portrait file;

[0027] The probability of an anomaly in the portrait archive is determined based on the time interval since the last archive anomaly detection, the first similarity and / or the second similarity, and the historical anomaly probability.

[0028] Optionally, in one embodiment of this application, determining a first similarity based on the similarity between the at least one feature image and the centroid, and / or determining a second similarity based on the similarity between the at least one feature image and the centroid, includes:

[0029] Obtain at least one face feature image and at least one human body feature image from the at least one feature image;

[0030] A first facial similarity is determined based on the similarity between the at least one facial feature image and the centroid of the face, and / or a second facial similarity is determined based on the similarity between the at least one facial feature image;

[0031] A first similarity of the human body is determined based on the similarity between the at least one human feature image and the centroid of the human body, and / or a second similarity of the human body is determined based on the similarity between the at least one human feature image;

[0032] According to preset rules, the smaller of the first similarity of the face and the first similarity of the body is selected as the first similarity, and the smaller of the second similarity of the face and the second similarity of the body is selected as the second similarity.

[0033] Secondly, this application also proposes an archival anomaly detection device, the device comprising:

[0034] The feature image acquisition module acquires at least one feature image of the portrait file, wherein the feature image includes a face image and / or a human body image whose similarity to the centroid of the portrait file is greater than a preset similarity threshold;

[0035] A similarity determination module is used to determine a first similarity based on the similarity between the at least one feature image and the centroid, and / or to determine a second similarity based on the similarity between the at least one feature image and the centroid.

[0036] An anomaly probability determination module is used to determine the anomaly probability of the portrait file based on the time interval since the last file anomaly detection, the first similarity and / or the second similarity.

[0037] Thirdly, this application also proposes a processing device, including a memory and a processor, wherein the memory stores computer program instructions, and the processor executes the computer program instructions to implement the steps of the above-described file anomaly detection method.

[0038] Fourthly, this application also proposes a computer-readable storage medium storing computer program instructions thereon, which, when executed by a processor, implement the steps of the above-described file anomaly detection method.

[0039] Fifthly, this application also proposes a chip including at least one processor, the processor being configured to execute computer program instructions stored in a memory to perform the steps of the above-described file anomaly detection method. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of this application, 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 this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This application provides an illustration of an application scenario for detecting file anomalies, as shown in one embodiment of this application.

[0042] Figure 2 This application provides a schematic flowchart of a method for detecting archival anomalies in one embodiment.

[0043] Figure 3 A schematic diagram of the module structure of an anomaly detection device 300 provided in one embodiment of this application;

[0044] Figure 4 This is a schematic diagram of the module structure of a processing device 400 provided in one embodiment of this application. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this application clearer, the application is described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the application. All other embodiments obtained by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application. Furthermore, it is understood that although the efforts made in such a development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, modifications to design, manufacturing, or production based on the technical content disclosed in this application are merely conventional technical means and should not be construed as insufficient disclosure of the content of this application.

[0046] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that is mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0047] Unless otherwise defined, the technical or scientific terms used in this application shall have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms "a," "an," "a kind," "the," and similar words used in this application do not indicate quantity limitation and may indicate singular or plural. The terms "comprising," "including," "having," and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units not listed, or may include other steps or units inherent to these processes, methods, products, or devices. The terms "connected," "linked," "coupled," and similar words used in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "A plurality" used in this application means two or more. The terms "first," "second," "third," etc., used in this application are merely to distinguish similar objects and do not represent a specific ordering of objects.

[0048] Furthermore, to better illustrate this application, numerous specific details are provided in the following detailed description. Those skilled in the art should understand that this application can be implemented without certain specific details. In some instances, apparatus, means, elements, and circuits well known to those skilled in the art have not been described in detail in order to highlight the main points of this application.

[0049] To clearly illustrate the technical solutions of the various embodiments of this application, the following describes... Figure 1 One exemplary scenario of an embodiment of this application will be described.

[0050] Please see Figure 1 , Figure 1This is a schematic diagram of a file anomaly detection system provided in an embodiment of this application. The system includes a data acquisition device 101, a clustering device 103, and a file anomaly detection device 105, wherein the data acquisition device 101, the clustering device 103, and the file anomaly detection device 105 can communicate with each other. The data acquisition device 101 can be an electronic device with data acquisition and data transmission / reception capabilities. For example, the data acquisition device 101 can include an electronic device capable of acquiring image information and / or video information of a target area, such as a camera, a lidar, etc. The camera can include a monocular camera, an infrared camera, a multi-view camera, a depth camera, etc. The lidar can include a single-line lidar, a multi-line lidar, etc., and this application does not impose any limitations. The data acquisition device 101 can send the acquired images to be clustered to the clustering device 103. The clustering device 103 can use clustering methods to cluster the acquired images to be clustered, calculate the similarity between images, and cluster images with high similarity into a portrait file. The clustering method can include K-means clustering, hierarchical clustering, etc. After clustering is completed, the clustering device 103 can send multiple portrait files to the file anomaly detection device 105, which then performs anomaly detection on the multiple portrait files. The file anomaly detection device 105 can be an electronic device with data processing and data transmission / reception capabilities, such as a physical device like a host, rack server, or blade server, or a virtual device like a virtual machine or container. It should be noted that the file anomaly detection device 105 can also be coupled internally to the portrait clustering device 103, enabling the portrait clustering device 103 to detect whether the multiple portrait files are abnormal.

[0051] The method for detecting archival anomalies described in this application will be explained in detail below with reference to the accompanying drawings. Figure 2 This is a schematic flowchart of one embodiment of the file anomaly detection method provided in this application. Although this application provides method operation steps as shown in the following embodiments or figures, more or fewer operation steps may be included in the method based on conventional or non-inventive effort. For steps that do not have a logically necessary causal relationship, the execution order of these steps is not limited to the execution order provided in the embodiments of this application. In actual file anomaly detection processes or when the device executes the method, it can be executed in the order shown in the embodiments or figures, or in parallel (e.g., in a parallel processor or multi-threaded processing environment).

[0052] Specifically, one implementation of the document anomaly detection method provided in this application is, for example... Figure 2 As shown, the method may include:

[0053] Step 201: Obtain at least one feature image of the portrait file, the feature image including a face image and / or a human body image whose similarity to the centroid of the portrait file is greater than a preset similarity threshold.

[0054] In this embodiment, the portrait profile can be obtained by clustering the images to be clustered by the clustering device 103. After clustering, a unique identifier can be assigned to the portrait profile, such as "Zhang San," "Li Si," etc. The portrait profile may include face images, body images, and face-body images. The face image may be a partial image extracted from the images to be clustered. After obtaining the portrait profile, the centroid of the portrait profile can be determined. The centroid can be a cluster centroid obtained by clustering using a clustering method, such as K-means clustering or hierarchical clustering. When the portrait profile includes both face images and body images, the centroid can be divided into face centroid and body centroid. After determining the centroid, at least one feature image can be determined based on the similarity between the images in the portrait profile and the centroid. In some embodiments of this application, the algorithm for determining the similarity may include a distance calculation method, a cosine similarity calculation method, or a kernel function calculation method. The distance calculation method may include Minkowski distance, Euclidean distance, Mahalanobis distance, etc. In one embodiment of this application, the smaller the distance between the image in the portrait file and the centroid, the higher the similarity between the image and the centroid. In one embodiment of this application, an image with a similarity greater than a preset similarity threshold to the centroid of the portrait file can be selected as the feature image; alternatively, the image with the highest similarity to the centroid of the portrait file can be selected as the feature image. It should be noted that when the portrait file includes both face images and body images, the feature images can be divided into two types: feature images for face images and feature images for body images. The feature image for face images can be determined based on the similarity between the face image in the portrait file and the face centroid, and the feature image for body images can be determined based on the similarity between the body image in the portrait file and the body centroid.

[0055] In this embodiment, since the feature image is a face image and / or human body image whose similarity to the centroid of the portrait file is greater than a preset similarity threshold, it can represent the most feature information of the portrait file, thereby providing a more accurate benchmark for subsequent similarity determination.

[0056] Step 203: Determine a first similarity based on the similarity between the at least one feature image and the centroid, and / or determine a second similarity based on the similarity between the at least one feature image and the centroid.

[0057] In this embodiment, a first similarity can be determined based on multiple similarities between the at least one feature image and the centroid. A second similarity can be determined based on the similarity between the at least one feature image and the centroid. In one embodiment, the similarity between feature images can be determined using methods such as perceptual hashing, histogram matching, and calculating normalized correlation coefficients. Specifically, in one embodiment, the first similarity can be determined based on the average of the multiple similarities. In another embodiment, a similarity greater than a preset threshold can be selected as the first similarity; this application does not impose any limitations on this. The method for determining the second similarity is the same as the method for determining the first similarity, and will not be described again here.

[0058] It should be noted that the number of feature images can be one, two, or more. When there is only one feature image, the second similarity is the similarity between the feature image and itself, i.e., the second similarity is 100%.

[0059] Furthermore, to improve the accuracy of the first similarity and / or the second similarity, in one embodiment of this application, determining the first similarity based on the similarity between the at least one feature image and the centroid, and / or determining the second similarity based on the similarity between the at least one feature image and the centroid, may include:

[0060] Step 301: Determine multiple first candidate similarities between the at least one feature image and the centroid, and determine the first candidate similarity that meets the first preset requirement as the first similarity; the first preset requirement includes the smallest similarity among the multiple first candidate similarities;

[0061] Step 303: and / or, determine the second candidate similarity between the at least one feature image, and determine the second candidate similarity less than the second preset threshold as the second similarity; the second preset requirement includes the smallest similarity among the plurality of second candidate similarities.

[0062] In this embodiment, multiple first candidate similarities between the at least one feature image and the centroid can be determined. If a first candidate similarity satisfies a first preset requirement, the first candidate similarity can be determined as the first similarity. Specifically, in one embodiment of this application, satisfying the first preset requirement may include the first similarity being the minimum among the multiple first candidate similarities, or it may include the first similarity being greater than a first preset similarity threshold. The first preset similarity threshold can be set by the user according to actual application needs, for example, it can be set to 99%, 98%, etc. In other embodiments of this application, multiple first candidate similarities greater than the first preset similarity threshold can be selected, and then the average value of the multiple first candidate similarities can be calculated, and the average value is determined as the first similarity. This application does not limit the first preset requirement here. The method for determining the second similarity can refer to the method for determining the first similarity, and will not be repeated here.

[0063] Step 205: Determine the probability of an anomaly in the portrait file based on the time interval since the last file anomaly detection, the first similarity and / or the second similarity.

[0064] In this embodiment, the time interval since the last anomaly detection can be determined by the difference between the time of the most recent anomaly detection of the portrait file and the time of the current anomaly detection. For example, the time interval can be 48 hours, 36 hours, etc. Anomalies in the portrait file can include images with other identifiers, such as clustering images belonging to portrait file B into portrait file A. The anomaly probability can be used to represent the likelihood of an anomaly in the portrait file; for example, an anomaly probability of 80% indicates that the portrait file is likely to be anomaly. In practical applications, new images are continuously added to the portrait file during the portrait clustering process. This leads to a longer time interval since the last anomaly detection, resulting in more new images being added and a higher probability of anomalies appearing in the portrait file. Therefore, in one embodiment of this application, the anomaly probability of the portrait file can be determined based on the time interval since the last anomaly detection.

[0065] Furthermore, since the first similarity is obtained by calculating the similarity between the at least one feature image and the centroid, if the first similarity is less than a set threshold, it indicates that the feature image has a low degree of similarity with other images, meaning that the portrait file is abnormal. Therefore, in one embodiment of this application, the first similarity can also be used to determine the probability of abnormality of the portrait file. The second similarity is similar to the first similarity; therefore, the probability of abnormality of the portrait file can also be determined based on the second similarity. Specifically, the probability of abnormality can be determined based on the time interval since the last file anomaly detection, the first similarity, and / or the second similarity and the probability of abnormality, and the preset correlation relationship can be represented by a correlation function, a correlation table, a correlation image, etc.

[0066] Through the above embodiments, during the anomaly detection process of portrait files, the probability of an anomaly in the portrait file can be determined by utilizing the similarity between the centroid of the portrait file and the feature image, as well as the time interval since the last anomaly detection. Therefore, it is not necessary to extract all information from the portrait file to determine whether it is abnormal, thereby improving the efficiency of anomaly detection and saving a significant amount of detection time.

[0067] Furthermore, in one embodiment of this application, determining the probability of an anomaly in the portrait file based on the time interval since the last file anomaly detection, the first similarity, and / or the second similarity may include:

[0068] Step 401: Determine the probability of an anomaly in the portrait file based on the positive correlation between the time interval since the last anomaly detection and the probability of anomaly in the portrait file, and the negative correlation between the first similarity and / or the second similarity and the probability of anomaly in the portrait file.

[0069] In this embodiment, the time interval since the last anomaly detection and the anomaly probability of the portrait file can be positively correlated. Here, "positive correlation" means that as the time interval since the last anomaly detection increases, the anomaly probability increases. In one example, the positive correlation may include a positive correlation coefficient between the time interval since the last anomaly detection and the anomaly probability. Alternatively, the first similarity and / or the second similarity may be negatively correlated with the anomaly probability of the portrait file. "Positive correlation" means that as the first similarity and / or the second similarity increases, the anomaly probability decreases. In one example, the negative correlation may include a negative correlation coefficient between the first similarity and / or the second similarity and the anomaly probability. Specifically, in one embodiment of this application, the anomaly probability P can be determined based on the correlation between the anomaly probability P, the time interval T, and the first similarity M, P = T / M. In another embodiment of this application, the anomaly probability P can also be determined based on the correlation between the anomaly probability P, the time interval T, and the second similarity N, P = T / N. Of course, in other embodiments of this application, the abnormal probability P of the human face can also be determined based on the correlation between the abnormal probability P and the time interval T, the first similarity M and the second similarity N, P = T / (M+N), and this application does not impose any restrictions on this.

[0070] Furthermore, to further improve the accuracy of detecting the portrait archives, in one embodiment of this application, determining the anomaly probability of the portrait archive based on the time interval since the last archive anomaly detection, the first similarity, and / or the second similarity may include:

[0071] Step 501: Determine the weighted value of the first similarity and the second similarity;

[0072] Step 503: Determine the probability of an anomaly in the portrait file based on the ratio of the weighted value to the time interval since the last file anomaly detection.

[0073] In this embodiment of the application, after determining the first similarity and the second similarity, the anomaly probability of the portrait file can be calculated according to the following formula:

[0074]

[0075] Wherein, α and β are the weighting coefficients of the first similarity and the second similarity. These weighting coefficients can be used to characterize the importance of the first and second similarities in influencing the anomaly probability. For example, the weighting coefficient β of the second similarity can be used to represent the impact of changes in the second similarity on the anomaly probability when the time interval and the first similarity remain unchanged. The weighting coefficients can be set by the user according to actual application needs, or determined based on past anomaly detection data. After determining the weighting coefficients, the weighted value of the first and second similarities can be determined. It should be noted that after determining the weighting coefficients of the first and second similarities, the weighting coefficient of the time interval can also be determined, and the formula is:

[0076] Through the above embodiments, the anomaly probability can be determined based on the ratio of the weighted value of the first similarity and the second similarity to the time interval, thereby enhancing the accuracy of anomaly file judgment.

[0077] In practical applications, to further improve the efficiency of anomaly detection in portrait archives, only a portion of the portrait archives that meet the requirements can be re-archived, thereby saving resources and improving the accuracy of the portrait archives. Based on this, in one embodiment of this application, after determining the anomaly probability of the portrait archive based on the time interval since the last anomaly detection, the first similarity, and / or the second similarity, the method may further include: re-archiving portrait archives whose anomaly probability is greater than a preset anomaly probability threshold.

[0078] In this embodiment, after obtaining the anomaly probability of the portrait file, the anomaly probability can be compared with a preset anomaly probability threshold. If the anomaly probability of the portrait file is greater than the preset anomaly probability threshold, the portrait file can be re-archived. Re-archiving may include performing file detection on the portrait file to detect and remove abnormal images from the portrait file, thereby correcting the abnormal files. Abnormal images may include images that should not belong to the portrait file, or images with poor image quality. In this embodiment, after removing the abnormal images, image quality checks can be performed on the abnormal images. If the abnormal images meet the image quality requirements, they can be returned to the data source for re-clustering, thereby classifying the abnormal images into new portrait files. If the abnormal images do not meet the image quality requirements, they can be directly deleted. Meeting the image quality requirements may include an image sharpness greater than a preset sharpness threshold or the number of feature values ​​contained in the image being higher than a quantity threshold.

[0079] Through the above embodiments, only the portrait files that meet the requirements can be re-archived, instead of re-archiving all portrait files. Compared with the existing technology that performs error checking on all portrait files at the same time, it can find files that may have errors more quickly and accurately, saving a lot of computing time.

[0080] Furthermore, in one embodiment of this application, the re-archiving of portrait files with an anomaly probability greater than a preset anomaly probability threshold may include:

[0081] Step 601: Determine at least one target portrait file whose anomaly probability is greater than a preset anomaly probability threshold;

[0082] Step 603: Determine the confidence level of each of the at least one target portrait file, wherein the confidence level is used to characterize the degree of correlation between the human body image in the target portrait file and the face image in the portrait file;

[0083] Step 605: Prioritize re-archiving target portrait files with confidence levels lower than the preset confidence threshold.

[0084] In this embodiment, at least one target portrait file with an anomaly probability greater than a preset anomaly probability threshold can be identified. In the actual portrait clustering process, the acquired images to be clustered can be processed to obtain face images. Then, all face images can be clustered to form multiple portrait files consisting solely of face images. If a human image in the images to be clustered has the same identifier as a face image in a certain portrait file, then the human image is added to that portrait file to form a face-body association file. Finally, all human images are matched with the images in the multiple portrait files. If a match is successful, the human image can be added to that portrait file to form a face-body strong association file. Based on this, the confidence level of the human image can be used to distinguish different types of portrait files. The confidence level can be used to represent the degree of association between the human image and the face image in the portrait file; that is, the higher the confidence level, the higher the degree of association between the corresponding human image and the face image. For example, the confidence level of the human body image in the pure face image portrait file can be 100%, the confidence level of the human body image in the face-body association file can be 60%, and the confidence level of the human body image in the face-body strong association file can be 40%. In one embodiment of this application, the confidence level of the at least one target portrait file can be determined respectively, and if the confidence level is less than a preset confidence threshold, the target portrait file is preferentially re-archived.

[0085] It should be noted that during the clustering process, high similarity between two captured images of different individuals may lead to anomalies in the pure face image profiles. Furthermore, human images may be associated with abnormal face images in the pure face image profiles, resulting in anomalies. Additionally, during image matching of all human images with images from the multiple profiles, high similarity between human images of different individuals may lead to matching errors. Based on the varying probabilities of anomalies in different types of profiles, if the anomaly probabilities of multiple target profiles are not significantly different, they can be re-archived sequentially according to their confidence levels. For example, in one instance, the anomaly probability of target profile 1 is 90.51%, that of target profile 2 is 90.57%, and that of target profile 3 is 90.61%. When the confidence level of target portrait file 1 is 100%, the confidence level of target portrait file 2 is 60%, and the confidence level of target portrait file 3 is 40%, target portrait file 3 is re-archived first, and then target portrait file 2 and target portrait file 3 are re-archived.

[0086] In practical applications, if a portrait file has been identified as an anomalous file in previous anomaly detections, the probability of it being an anomalous file in the current anomaly detection increases. Therefore, in one embodiment of this application, determining the anomalous probability of the portrait file based on the time interval since the last anomaly detection, the first similarity, and / or the second similarity may include:

[0087] Step 701: Obtain the historical anomaly probability of the portrait file;

[0088] Step 703: Determine the anomaly probability of the portrait archive based on the time interval since the last archive anomaly detection, the first similarity and / or the second similarity, and the historical anomaly probability.

[0089] In this embodiment of the application, to ensure the accuracy of clustering, anomaly detection can be performed periodically on the portrait files obtained after clustering. Based on this, multiple anomaly detection results of the portrait files can be obtained, thereby determining the historical anomaly probability of the portrait files. After determining the anomaly probability, the anomaly probability of the portrait files can be determined based on the time interval since the last anomaly detection, the first similarity and / or the second similarity, and the historical anomaly probability. Specifically, in one embodiment of this application, the anomaly probability of the portrait files can be calculated according to the following formula:

[0090]

[0091] The x value can be determined based on the historical anomaly probability. For example, if the historical anomaly probability is 0.3, the x value can be 0.3.

[0092] Through the above embodiments, the historical anomaly probability of the portrait file can be determined based on multiple anomaly detections, thereby providing assistance in determining the current anomaly probability and determining the anomaly probability of the portrait file more accurately and quickly.

[0093] Furthermore, in one embodiment of this application, determining a first similarity based on the similarity between the at least one feature image and the centroid, and / or determining a second similarity based on the similarity between the at least one feature image and the centroid, may include:

[0094] Step 801: Obtain at least one face feature image and at least one human body feature image from the at least one feature image;

[0095] Step 803: Determine a first facial similarity based on the similarity between the at least one facial feature image and the centroid of the face, and / or determine a second facial similarity based on the similarity between the at least one facial feature image;

[0096] Step 805: Determine a first similarity of the human body based on the similarity between the at least one human feature image and the centroid of the human body, and / or determine a second similarity of the human body based on the similarity between the at least one human feature image;

[0097] Step 807: According to preset rules, select the smaller of the first similarity of the face and the first similarity of the human body as the first similarity, and select the smaller of the second similarity of the face and the second similarity of the human body as the second similarity.

[0098] In this embodiment, when the portrait file includes both face images and body images, since the centroids of the face images and body images are not the same, there are two centroids in the portrait file: the face centroid and the body centroid. Correspondingly, the feature images can be divided into face feature images and body feature images. Following the similarity determination method described above, the first similarity and second similarity of the face images in the portrait file, and the first similarity and second similarity of the body images, are determined sequentially. In one embodiment of this application, to improve the accuracy of anomaly detection, the portrait file can be judged as abnormal based on the smaller similarity. Therefore, the smaller of the first face similarity and the first body similarity can be selected as the first similarity, and the smaller of the second face similarity and the second body similarity can be selected as the second similarity. Correspondingly, the anomaly probability can be determined according to the following formula:

[0099]

[0100] Wherein, M1 and M2 represent the first and second similarity scores of the faces, and N1 and N2 represent the first and second similarity scores of the human body. The min() function can be used to find the minimum value in a set of numbers.

[0101] Through the above embodiments, anomaly detection can be performed on portrait archives that contain both facial images and body images, thereby making the archive anomaly detection method applicable to different portrait archives and expanding the scope of application of archive anomaly detection.

[0102] This application also provides a file anomaly detection device 105, such as... Figure 3 As shown, the device 105 may include:

[0103] The feature image acquisition module 301 is used for at least one feature image of the portrait file, the feature image including a face image and / or a human body image whose similarity to the centroid of the portrait file is greater than a first preset threshold.

[0104] The similarity determination module 303 is used to determine a first similarity based on the similarity between the at least one feature image and the centroid, and / or to determine a second similarity based on the similarity between the at least one feature image and the centroid.

[0105] The anomaly probability determination module 305 is used to determine the anomaly probability of the portrait file based on the time interval since the last file anomaly detection, the first similarity and / or the second similarity.

[0106] Optionally, in one embodiment of this application, determining a first similarity based on the similarity between the at least one feature image and the centroid, and / or determining a second similarity based on the similarity between the at least one feature image and the centroid, includes:

[0107] Multiple first candidate similarities between the at least one feature image and the centroid are determined, and the first candidate similarity that meets the first preset requirement is determined as the first similarity; the first preset requirement includes the smallest similarity among the multiple first candidate similarities;

[0108] And / or, determine the second candidate similarity between the at least one feature image, and determine the second candidate similarity less than the second preset threshold as the second similarity; the second preset requirement includes the smallest similarity among the plurality of second candidate similarities.

[0109] Optionally, in one embodiment of this application, determining the anomaly probability of the portrait file based on the time interval since the last file anomaly detection, the first similarity, and / or the second similarity includes:

[0110] The probability of an anomaly in the portrait file is determined based on the positive correlation between the time interval since the last anomaly detection and the probability of anomaly in the portrait file, and the negative correlation between the first similarity and / or the second similarity and the probability of anomaly in the portrait file.

[0111] Optionally, in one embodiment of this application, determining the anomaly probability of the portrait file based on the positive correlation between the time interval since the last anomaly detection and the anomaly probability of the portrait file, and the negative correlation between the first similarity and / or the second similarity and the anomaly probability of the portrait file, includes:

[0112] Determine the weighted value of the first similarity and the second similarity;

[0113] The probability of an anomaly in the portrait file is determined by the ratio of the weighted value to the time interval since the last anomaly detection.

[0114] Optionally, in one embodiment of this application, after determining the anomaly probability of the portrait file based on the time interval since the last file anomaly detection, the first similarity, and / or the second similarity, the method further includes:

[0115] The portrait files with an anomaly probability greater than a preset anomaly probability threshold are re-archived.

[0116] Optionally, in one embodiment of this application, the step of re-archiving the portrait files whose anomaly probability is greater than a preset anomaly probability threshold includes:

[0117] Identify at least one target portrait file whose anomaly probability is greater than a preset anomaly probability threshold;

[0118] The confidence level of each of the at least one target portrait file is determined, wherein the confidence level is used to characterize the degree of correlation between the human body image in the target portrait file and the face image in the portrait file;

[0119] Target portrait files with a confidence level lower than a preset confidence threshold are preferentially re-archived.

[0120] Optionally, in one embodiment of this application, determining the anomaly probability of the portrait file based on the time interval since the last file anomaly detection, the first similarity, and / or the second similarity includes:

[0121] Obtain the historical anomaly probability of the aforementioned portrait file;

[0122] The probability of an anomaly in the portrait archive is determined based on the time interval since the last archive anomaly detection, the first similarity and / or the second similarity, and the historical anomaly probability.

[0123] Optionally, in one embodiment of this application, determining a first similarity based on the similarity between the at least one feature image and the centroid, and / or determining a second similarity based on the similarity between the at least one feature image and the centroid, includes:

[0124] Obtain at least one face feature image and at least one human body feature image from the at least one feature image;

[0125] A first facial similarity is determined based on the similarity between the at least one facial feature image and the centroid of the face, and / or a second facial similarity is determined based on the similarity between the at least one facial feature image;

[0126] A first similarity of the human body is determined based on the similarity between the at least one human feature image and the centroid of the human body, and / or a second similarity of the human body is determined based on the similarity between the at least one human feature image;

[0127] According to preset rules, the smaller of the first similarity of the face and the first similarity of the body is selected as the first similarity, and the smaller of the second similarity of the face and the second similarity of the body is selected as the second similarity.

[0128] This application also provides a processing device, including a memory and a processor, wherein the memory stores computer program instructions, and the processor is configured to run the computer program instructions to perform the file anomaly detection method described in the above embodiments.

[0129] The processing device can be a physical device or a cluster of physical devices, or it can be a virtualized cloud device, such as at least one cloud computing device in a cloud computing cluster. For ease of understanding, this application illustrates the structure of the processing device as an independent physical device.

[0130] like Figure 4 As shown, the processing device 400 includes a processor and a memory for storing computer program instructions for the processor; wherein the processor is configured to implement the above-described means when executing the computer program instructions. The processing device 400 includes a memory 401, a processor 403, a bus 405, and a communication interface 407. The memory 401, processor 403, and communication interface 407 communicate via the bus 405. The bus 405 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, Figure 4 The symbol is represented by only one thick line, but this does not indicate that there is only one bus or one type of bus. Communication interface 407 is used for communication with external devices.

[0131] The processor 403 can be a central processing unit (CPU). The memory 401 can include volatile memory, such as random access memory (RAM). The memory 401 can also include non-volatile memory, such as read-only memory (ROM), flash memory, HDD or SSD, etc.

[0132] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0133] This application also provides a chip including at least one processor, the processor being configured to run computer program instructions stored in a memory to perform the steps of the file anomaly detection method described in the above embodiments.

[0134] This application also provides a computer-readable storage medium storing computer program instructions thereon, which, when executed by a processor, implement the steps of the above-described file anomaly detection method.

[0135] A computer-readable storage medium can be a tangible device capable of holding and storing instructions for use by an instruction execution device. A computer-readable storage medium can be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), electrically programmable read-only memory (EPROM or flash memory), static random-access memory (SRAM), compact disc read-only memory (CD-ROM), digital video disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination thereof.

[0136] The computer program instructions described herein can be downloaded from a computer-readable storage medium to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper cables, fiber optic cables, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer program instructions from the network and forwards them to a computer-readable storage medium within the respective computing / processing device.

[0137] The computer program instructions used to perform the operations of this application may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as "C" or similar languages. The computer program instructions may be executed entirely on the user's computer, partially on the user's computer, as a standalone 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 remote computers, the remote computer may be connected to the user's computer via 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., via the Internet using an Internet service provider). In some embodiments, electronic circuits, such as programmable logic circuits, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), are personalized by utilizing state information from computer program instructions. These electronic circuits can execute computer program instructions to implement various aspects of this application.

[0138] Various aspects of this application are described herein with reference to flowchart illustrations and / or block diagrams of methods and apparatus according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions.

[0139] These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0140] Computer program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other equipment to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other equipment to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other equipment to perform the functions / actions specified in one or more blocks of a flowchart and / or block diagram.

[0141] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatuses, systems, and methods according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved.

[0142] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for detecting archival anomalies, characterized in that, The method includes: Acquire at least one feature image of the portrait profile, the feature image including a face image and / or a human body image whose similarity to the centroid of the portrait profile is greater than a preset similarity threshold; A first similarity is determined based on the similarity between the at least one feature image and the centroid, and / or a second similarity is determined based on the similarity between the at least one feature image and the centroid. The probability of an anomaly in the portrait archive is determined based on the time interval since the last archive anomaly detection, the first similarity and / or the second similarity. Determining the probability of an anomaly in the portrait archive based on the time interval since the last archive anomaly detection, the first similarity, and / or the second similarity includes: The probability of an anomaly in the portrait file is determined based on the positive correlation between the time interval since the last anomaly detection and the probability of anomaly in the portrait file, and the negative correlation between the first similarity and / or the second similarity and the probability of anomaly in the portrait file. The step of determining the anomaly probability of a portrait file based on the positive correlation between the time interval since the last anomaly detection and the anomaly probability of the portrait file, and the negative correlation between the first similarity and / or the second similarity and the anomaly probability of the portrait file, includes: Determine the weighted value of the first similarity and the second similarity; The probability of an anomaly in the portrait file is determined by the ratio of the weighted value to the time interval since the last anomaly detection.

2. The method according to claim 1, characterized in that, Determining a first similarity based on the similarity between the at least one feature image and the centroid, and / or determining a second similarity based on the similarity between the at least one feature image and the centroid, includes: Multiple first candidate similarities between the at least one feature image and the centroid are determined, and the first candidate similarity that meets the first preset requirement is determined as the first similarity; the first preset requirement includes the smallest similarity among the multiple first candidate similarities; And / or, determine the second candidate similarity between the at least one feature image, and determine the second candidate similarity that meets the second preset requirement as the second similarity; the second preset requirement includes the smallest similarity among the plurality of second candidate similarities.

3. The method according to claim 1, characterized in that, After determining the probability of an anomaly in the portrait file based on the time interval since the last anomaly detection, the first similarity, and / or the second similarity, the method further includes: The portrait files with an anomaly probability greater than a preset anomaly probability threshold are re-archived.

4. The method according to claim 3, characterized in that, The re-archiving of portrait files with an anomaly probability greater than a preset anomaly probability threshold includes: Identify at least one target portrait file whose anomaly probability is greater than a preset anomaly probability threshold; The confidence level of each of the at least one target portrait file is determined, wherein the confidence level is used to characterize the degree of correlation between the human body image in the target portrait file and the face image in the portrait file; Target portrait files with a confidence level lower than a preset confidence threshold are preferentially re-archived.

5. The method according to claim 1, characterized in that, Determining the probability of an anomaly in the portrait archive based on the time interval since the last archive anomaly detection, the first similarity, and / or the second similarity includes: Obtain the historical anomaly probability of the aforementioned portrait file; The probability of an anomaly in the portrait archive is determined based on the time interval since the last archive anomaly detection, the first similarity and / or the second similarity, and the historical anomaly probability.

6. The method according to claim 1, characterized in that, Determining a first similarity based on the similarity between the at least one feature image and the centroid, and / or determining a second similarity based on the similarity between the at least one feature image and the centroid, includes: Obtain at least one face feature image and at least one human body feature image from the at least one feature image; A first facial similarity is determined based on the similarity between the at least one facial feature image and the centroid of the face, and / or a second facial similarity is determined based on the similarity between the at least one facial feature image; A first similarity of the human body is determined based on the similarity between the at least one human feature image and the centroid of the human body, and / or a second similarity of the human body is determined based on the similarity between the at least one human feature image; According to preset rules, the smaller of the first similarity of the face and the first similarity of the body is selected as the first similarity, and the smaller of the second similarity of the face and the second similarity of the body is selected as the second similarity.

7. A device for detecting archival anomalies, characterized in that, The device includes: The feature image acquisition module acquires at least one feature image of the portrait file, wherein the feature image includes a face image and / or a human body image whose similarity to the centroid of the portrait file is greater than a preset similarity threshold; A similarity determination module is used to determine a first similarity based on the similarity between the at least one feature image and the centroid, and / or to determine a second similarity based on the similarity between the at least one feature image and the centroid. An anomaly probability determination module is used to determine the anomaly probability of the portrait archive based on the time interval since the last archive anomaly detection, the first similarity and / or the second similarity; The anomaly probability determination module is further configured to determine the anomaly probability of the portrait file based on the positive correlation between the time interval since the last anomaly detection and the anomaly probability of the portrait file, and the negative correlation between the first similarity and / or the second similarity and the anomaly probability of the portrait file. The anomaly probability determination module is further configured to determine the weighted value of the first similarity and the second similarity; and to determine the anomaly probability of the portrait archive based on the ratio of the weighted value to the time interval from the last archive anomaly detection.

8. A processing device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the file anomaly detection method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the steps of the file anomaly detection method according to any one of claims 1 to 6.

10. A chip, characterized in that, It includes at least one processor, which is configured to execute computer program instructions stored in a memory to perform the steps of the file anomaly detection method according to any one of claims 1 to 6.

Images