A method, apparatus, terminal and computer-readable storage medium for document aggregation evaluation

Abstract

This invention provides a clustering evaluation method, apparatus, terminal, and computer-readable storage medium. The clustering evaluation method includes: acquiring a clustered image set; determining a first motion trajectory of a target object based on the spatiotemporal information of each clustered image; comparing the first motion trajectory of the target object with the first motion trajectory of another target to determine whether the target is a peer object of the target object; the first motion trajectory includes at least two positions; in response to the target being a peer object of the target object, determining the clustering evaluation result of the clustered image set based on the first motion trajectories corresponding to the target object and its peer objects. This application, by determining the peer objects of the target object and determining the clustering evaluation result of the clustered image set based on the first motion trajectory of the target object and the first motion trajectories of all peer objects of the target object, can more comprehensively and accurately evaluate the clustering effect of the clustered image set.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method, apparatus, terminal, and computer-readable storage medium for image aggregation evaluation. Background Technology

[0002] With the continuous advancement of image recognition and video surveillance technologies, the application of big data is receiving increasing attention, especially facial recognition image aggregation, which occupies a crucial position within big data and its effects are becoming increasingly apparent. To verify the practicality of our products and identify areas for product optimization, we need to statistically analyze the image aggregation effects in actual production environments. However, existing statistical methods are either cumbersome, requiring significant manpower and resources, or lack comprehensive coverage and sufficient statistical dimensions. Summary of the Invention

[0003] The main technical problem solved by this invention is to provide an article detection method, terminal and computer-readable storage medium, which solves the problem of inaccurate evaluation of the aggregation effect in the prior art.

[0004] To solve the above-mentioned technical problems, the first technical solution adopted by the present invention is: to provide a clustering evaluation method, the clustering evaluation method comprising:

[0005] Acquire a clustered image set, which consists of multiple clustered images of the target object in different scenes; each clustered image contains spatiotemporal information.

[0006] Based on the spatiotemporal information of each image, the first motion trajectory of the target object is determined;

[0007] The first motion trajectory of the target object is compared with the first motion trajectory of another target to determine whether the target is a peer of the target object; the first motion trajectory includes at least two positions;

[0008] In response to a target object in the same row as the target object, the clustering evaluation result of the clustered image set is determined based on the first motion trajectory corresponding to the target object and the target object's corresponding object in the same row.

[0009] The first trajectory includes a starting point and an ending point.

[0010] Compare the first motion trajectory of the target object with the first motion trajectory of another target to determine whether the target is a peer of the target object, including:

[0011] Compare the start and end points of the target object with the start and end points of the target;

[0012] If the starting point and ending point of the target object and the target are the same, then it is determined whether the target is a peer object of the target object based on the points contained in the first motion trajectory of the target object and the points contained in the first motion trajectory of the target.

[0013] The determination of whether a target is a peer object of the target object based on the positions contained in the first motion trajectory of the target object and the positions contained in the first motion trajectory of the target includes:

[0014] The probability value of being in the same lane between a target object and a target is determined by the ratio of the number of intersection points to the number of union points between the points corresponding to the points corresponding to the first movement trajectory of the target object and the points corresponding to the first movement trajectory of the target object. The number of intersection points is the number of points corresponding to the intersection of the points in the first movement trajectory of the target object and the points in the first movement trajectory of the target object. The number of union points is the number of points corresponding to the union of the points in the first movement trajectory of the target object and the points in the first movement trajectory of the target object.

[0015] If the probability value of a peer exceeds the probability threshold, then the target is identified as a peer of the target object.

[0016] The clustering evaluation result of the clustered image set is determined based on the first motion trajectory corresponding to the target object and its peer objects, including:

[0017] Based on the first motion trajectory of the target object and the first motion trajectories corresponding to all objects in the same row as the target object, a second motion trajectory is determined that is common to both the target object and the objects in the same row; the position corresponding to the second motion trajectory is the union of all positions contained in the first motion trajectories corresponding to both the target object and the objects in the same row.

[0018] Based on the second motion trajectory, the first motion trajectory of the target object, and the first motion trajectories corresponding to all objects in the same row as the target object, the clustering evaluation result of the clustered image set is determined.

[0019] Before the step of determining the clustering evaluation result of the clustered image set based on the second motion trajectory, the first motion trajectory of the target object, and the first motion trajectories corresponding to all objects in the same row as the target object, the following steps are also included:

[0020] The corresponding points in the second motion trajectory are denoised using the clustered images acquired from adjacent points in the first motion trajectory of the target object and / or peer objects.

[0021] Specifically, the denoising process for corresponding points in the second motion trajectory is performed on the clustered images acquired based on adjacent points in the first motion trajectory of the target object and / or peer objects, including:

[0022] Traverse all points in the second motion trajectory and select one point as a candidate point;

[0023] Determine whether a candidate site exists only in a single first motion trajectory;

[0024] If a candidate site exists only in a first motion trajectory, the motion velocity corresponding to the candidate site is determined based on the spatiotemporal information of the aggregated image of the candidate site in the first motion trajectory and the spatiotemporal information of the aggregated image of the previous point of the candidate site.

[0025] If the movement speed of a candidate site exceeds a speed threshold, the candidate site is deleted.

[0026] The results of the clustering evaluation include the clustering accuracy rate;

[0027] Based on the second motion trajectory, the first motion trajectory of the target object, and the first motion trajectories corresponding to all objects in the same row as the target object, the clustering evaluation result of the clustered image set is determined, including:

[0028] Based on the second motion trajectory obtained after denoising, determine the number of clustered images collected for the noise sites corresponding to the target object and the object in the same row, respectively.

[0029] The clustering accuracy of the clustered image set is determined based on the number of clustered images collected from the noise sites corresponding to the target object and the objects in the same row.

[0030] The clustering accuracy of the clustered image set is determined by the number of clustered images collected based on the noise sites corresponding to the target object and the objects in the same row, including:

[0031] The number of noisy images is obtained by summing the number of images collected from the noise sites corresponding to the target object and all objects in the same row of the target object.

[0032] The aggregation accuracy of the aggregation image set is determined by the ratio of the number of noisy images to the total number of aggregation images corresponding to the target object and all its corresponding peer objects.

[0033] Among them, the cluster evaluation results include the cluster recall rate;

[0034] Based on the second motion trajectory, the first motion trajectory of the target object, and the first motion trajectories corresponding to all objects in the same row as the target object, the clustering evaluation result of the clustered image set is determined, including:

[0035] If a point in the second motion trajectory is a missed image location of the target object, then it is determined whether the target object has an uncollected image at the missed image location based on the archived image of the target object's peer at the missed image location.

[0036] Traverse the missed images of the target object and the objects in the same row to determine the unaggregated images corresponding to the target object and the objects in the same row, respectively.

[0037] The clustering recall rate of the clustered image set is determined based on the number of missed sites corresponding to all non-clustered images of the target object and the number of sites corresponding to all clustered images.

[0038] The spatiotemporal information includes the capture time.

[0039] Determining whether the target object has un-aggregated images at the missed imaging sites based on the archived images of peers of the target object at the missed imaging sites includes:

[0040] Select a composite image of a subject from the same field at a missed location;

[0041] Based on the capture time of the archived images of the selected peer objects at the missed capture sites, images captured at the missed capture sites within a preset time period are selected as candidate images.

[0042] Based on the similarity between each candidate image and an image containing the target object, it is determined whether the candidate image is an uncollected image of the target object collected at a missed capture site.

[0043] Specifically, based on the similarity between each candidate image and an image containing the target object, it is determined whether the candidate image is an uncollected image of the target object captured at a missed imaging site, including:

[0044] If the similarity between a candidate image and an image containing the target object exceeds a similarity threshold, the candidate image corresponding to the similarity threshold is retained.

[0045] If at least two candidate images are retained corresponding to the missed image site, the candidate image with the highest similarity is selected as the uncollected image of the target object at the missed image site.

[0046] The clustering recall rate of the clustered image set is determined based on the number of missed sites corresponding to all non-clustered images and the number of sites corresponding to all clustered images, respectively, including:

[0047] The number of sites corresponding to all clustered images of the target object and the number of sites corresponding to all clustered images of all objects in the same row of the target object are summed to determine the number of the first point.

[0048] The second number of sites is obtained by summing the number of sites corresponding to all archived images of the target object, the number of missed sites corresponding to unarchived images, and the number of sites corresponding to all archived images of all objects in the same row of the target object, and the number of missed sites corresponding to unarchived images.

[0049] The cluster recall rate of the clustered image set is determined based on the ratio of the number of first points to the number of second points.

[0050] To solve the above-mentioned technical problems, the second technical solution adopted by the present invention is: to provide a data aggregation evaluation device, the data aggregation evaluation device comprising:

[0051] The acquisition module is used to acquire a clustered image set, which consists of multiple clustered images of the target object in different scenes; each clustered image has spatiotemporal information.

[0052] The processing module is used to determine the first motion trajectory of the target object based on the spatiotemporal information of each image.

[0053] The comparison module is used to compare the first motion trajectory of the target object with the first motion trajectory of another target to determine whether the target is a peer of the target object; the first motion trajectory includes at least two positions;

[0054] The analysis module is used to determine the clustering evaluation result of the clustered image set based on the first motion trajectory corresponding to the target object and the target object's peer object, in response to the target object being a peer object.

[0055] To solve the above-mentioned technical problems, the third technical solution adopted by the present invention is to provide a terminal, the terminal including a memory, a processor and a computer program stored in the memory and running on the processor, the processor being used to execute program data to implement the steps in the above-described file evaluation method.

[0056] To solve the above-mentioned technical problems, the fourth technical solution adopted by the present invention is to provide a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps in the above-mentioned document evaluation method.

[0057] The beneficial effects of this invention are as follows: Unlike existing technologies, this invention provides a clustering evaluation method, apparatus, terminal, and computer-readable storage medium. The clustering evaluation method includes: acquiring a clustered image set, which consists of multiple clustered images of a target object captured in different scenes; each clustered image has spatiotemporal information; determining a first motion trajectory of the target object based on the spatiotemporal information of each clustered image; comparing the first motion trajectory of the target object with the first motion trajectory of another target to determine whether the target is a peer of the target object; the first motion trajectory includes at least two positions; and, in response to the target being a peer of the target object, determining the clustering evaluation result of the clustered image set based on the first motion trajectories corresponding to the target object and its peers. This application, by determining the peers of the target object and determining the clustering evaluation result of the clustered image set based on the first motion trajectory of the target object and the first motion trajectories of all peers of the target object, can more comprehensively and accurately evaluate the clustering effect of the clustered image set. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0059] Figure 1 This is a flowchart illustrating the data aggregation and evaluation method provided by the present invention;

[0060] Figure 2 yes Figure 1 A flowchart illustrating a specific embodiment of step S3 in the provided data aggregation and evaluation method;

[0061] Figure 3 yes Figure 2 A flowchart illustrating a specific embodiment of step S32 in the provided data aggregation and evaluation method;

[0062] Figure 4 This is a schematic diagram of the first motion trajectory of the target object X;

[0063] Figure 5 This is a schematic diagram of the first motion trajectory of target Y;

[0064] Figure 6 This is a schematic diagram of the first motion trajectory of target Z;

[0065] Figure 7 yes Figure 1 A flowchart illustrating a specific embodiment of step S4 in the provided data aggregation and evaluation method;

[0066] Figure 8 This is a schematic diagram of the second motion trajectory corresponding to the target object X and its peer objects Y and Z.

[0067] Figure 9 yes Figure 7 A flowchart illustrating a specific embodiment of step S42 in the provided data aggregation and evaluation method;

[0068] Figure 10 This is a schematic diagram of the second motion trajectory after noise reduction processing;

[0069] Figure 11 yes Figure 7 A flowchart illustrating a specific embodiment of step S43 in the provided data aggregation and evaluation method;

[0070] Figure 12 yes Figure 7 A flowchart illustrating another specific embodiment of step S43 in the provided data aggregation and evaluation method;

[0071] Figure 13 This is a schematic diagram of the framework of an embodiment of the data collection and evaluation device provided by the present invention;

[0072] Figure 14 This is a schematic diagram of the framework of an embodiment of the terminal provided by the present invention;

[0073] Figure 15 A schematic diagram of a framework of an embodiment of a computer-readable storage medium provided by the present invention. Detailed Implementation

[0074] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0075] In the following description, specific details such as particular system architectures, interfaces, and technologies are presented for illustrative purposes rather than for limiting purposes, in order to provide a thorough understanding of this application.

[0076] In this article, the term "and / or" simply describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / " generally indicates that the preceding and following related objects have an "or" relationship. Furthermore, "more" in this article means two or more objects.

[0077] To enable those skilled in the art to better understand the technical solution of the present invention, the following describes in further detail a data aggregation evaluation method provided by the present invention in conjunction with the accompanying drawings and specific embodiments.

[0078] Please see Figure 1 , Figure 1This is a flowchart illustrating the data aggregation evaluation method provided by the present invention.

[0079] This embodiment provides a clustering assessment method, which includes the following steps.

[0080] S1: Obtain the archived image set, which consists of multiple archived images of the target object in different scenes; each archived image has spatiotemporal information.

[0081] S2: Based on the spatiotemporal information of each image, determine the first motion trajectory of the target object.

[0082] S3: Compare the first motion trajectory of the target object with the first motion trajectory of another target to determine whether the target is a peer of the target object; the first motion trajectory includes at least two positions.

[0083] S4: In response to a target object being a peer object, determine the clustering evaluation result of the clustered image set based on the first motion trajectory corresponding to the target object and the peer object of the target object respectively.

[0084] Specifically, the steps for obtaining the aggregated image set in step S1 are as follows.

[0085] Images containing the target are acquired by image acquisition devices installed at preset locations within a preset area, resulting in images to be aggregated. In this embodiment, the target can be a person or a moving target such as an animal. In this embodiment, the images to be aggregated are images containing a person.

[0086] Deep learning algorithms are used to perform image recognition on the collected images to be aggregated, and the structured information of the images is obtained. For example, the structured information includes image ID, spatiotemporal information of the images to be aggregated, facial feature values, face width, rotation angle, face confidence score, whether the image is wearing a mask, hat, or glasses, and mode of transportation.

[0087] In this embodiment, the image acquisition devices installed at each preset location are considered independent sites.

[0088] The images to be clustered within a preset time period are clustered using a preset clustering algorithm to obtain the clustered image set corresponding to each target. The preset time period can be one day, one month, or 12 hours, and the specific preset time period is set according to the actual situation. For example, the preset clustering algorithm can be K-MEANS clustering algorithm, mean shift clustering algorithm, DBSCAN clustering algorithm, hierarchical clustering algorithm, etc.

[0089] In one embodiment, the feature similarity between facial feature values ​​of two images to be clustered is calculated. If the feature similarity is greater than a feature similarity threshold, the two images corresponding to the feature similarity are assigned to the same clustered image set. The feature similarity can be cosine similarity or Manhattan distance (L1).

[0090] In one specific embodiment, the cosine similarity between the facial feature values ​​of two images to be aggregated is calculated. The cosine similarity value ranges from [0, 1], with a larger value indicating greater similarity between the two images.

[0091]

[0092] In the formula: similarity represents the cosine similarity between the facial feature values ​​of two images to be aggregated; A i B i This represents the facial feature values ​​corresponding to the two images to be aggregated.

[0093] By calculating the cosine similarity between the facial feature values ​​of two images to be aggregated, the images to be aggregated at each point within a preset time period are aggregated to obtain the aggregated image set corresponding to each target.

[0094] Arbitrarily select a target as the target object. The archived image set of the target object consists of multiple archived images of the target object captured in different scenes. The spatiotemporal information of each archived image includes the capture time and capture position of the archived image.

[0095] Specifically, the steps for determining the first motion trajectory of the target object based on the spatiotemporal information of each image in step S2 are as follows.

[0096] The images in the aggregated image set are sorted based on their capture time. The first motion trajectory of the target object is determined according to the order of the capture points corresponding to the sorted images. These capture points can also be referred to as locations. World coordinates, i.e., the latitude and longitude coordinates of the location, can be determined based on the capture points.

[0097] Based on the capture time of each image in the target's image set, the images in the image set are sorted, and the first motion trajectory of the target is determined according to the order of the capture points corresponding to the sorted images.

[0098] In one embodiment, the first motion trajectory includes a start point and an end point. For example, the first motion trajectory consists of multiple points.

[0099] Specifically, the steps in step S3 to compare the first motion trajectory of the target object with the first motion trajectory of another target to determine whether the target is a peer of the target object are as follows.

[0100] Specifically, two targets, X and Y, are considered to be "co-targets" within a given time frame. Target X passes through m points and target Y passes through n points within the same time frame. If the ratio between the intersection of the m and n points and the union of the m and n points reaches a certain proportion (e.g., 60%), then X and Y are defined as co-targets.

[0101] Please see Figure 2 , Figure 2 yes Figure 1 A flowchart illustrating a specific embodiment of step S3 in the provided data aggregation and evaluation method.

[0102] In one embodiment, it is determined whether the target is a peer object of the target object by the following steps.

[0103] S31: Compare the start and end points of the target object with the start and end points of the target.

[0104] Specifically, the starting point of the target object is compared with the starting point of a target to determine whether the starting point of the target object is the same as the starting point of the target.

[0105] The end point of the target object is compared with the end point of the target to determine whether the end point of the target object is the same as the end point of the target.

[0106] S32: In response to the fact that the starting point and ending point of the target object and the target are the same, determine whether the target is a peer object of the target object based on the points contained in the first motion trajectory of the target object and the points contained in the first motion trajectory of the target.

[0107] Specifically, based on the following specific steps, it is determined whether the target is a peer of the target object according to the position contained in the first motion trajectory and the position contained in the first motion trajectory of the target.

[0108] Please see Figure 3 , Figure 3 yes Figure 2 A flowchart illustrating a specific embodiment of step S32 in the provided data collection and evaluation method.

[0109] S321: Determine the same-track probability value between the target object and the target based on the ratio of the number of intersection points to the number of union points between the points corresponding to the first movement trajectory of the target object and the points corresponding to the first movement trajectory of the target.

[0110] Specifically, the m positions corresponding to the first motion trajectory of the target object and the n positions corresponding to the first motion trajectory of the target are statistically analyzed.

[0111] The ratio between the number of intersection points of the m points corresponding to the first motion trajectory of the target object and the n points corresponding to the first motion trajectory of the target object is calculated and the number of union points of the m points and the n points is used as the same-line probability value between the target object and the target object.

[0112] The number of intersection points is the number of points corresponding to the intersection of points in the first motion trajectory of the target object and points in the first motion trajectory of the target; the number of union points is the number of points corresponding to the union of points in the first motion trajectory of the target object and points in the first motion trajectory of the target.

[0113] S322: In response to the peer probability value exceeding the probability threshold, the target is determined to be a peer object of the target object.

[0114] Please see Figures 4 to 6 , Figure 4 This is a schematic diagram of the first motion trajectory of the target object X; Figure 5 This is a schematic diagram of the first motion trajectory of target Y; Figure 6 This is a schematic diagram of the first motion trajectory of target Z.

[0115] In one specific embodiment, the positions corresponding to the first motion trajectory of target object X include A, C, D, E, G, H, I, J; the positions corresponding to the first motion trajectory of target Y include A, C, D, E, F, H, I, J; and the positions corresponding to the first motion trajectory of target Z include A, B, C, C', E, G, H, I, J.

[0116] The intersection of target object X and target Y is A, C, D, E, H, I, J, i.e., X∩Y={A, C, D, E, H, I, J}; the intersection of target object X and target Z is A, C, E, G, H, I, J, i.e., X∩Z={A, C, E, G, H, I, J}; the union of target object X and target Y is A, C, D, E, F, G, H, I, J, i.e., X∪Y={A, C, D, E, F, G, H, I, J}; and the union of target object X and target Z is A, B, C, C', E, F, G, H, I, J, i.e., X∪Z={A, B, C, C', E, F, G, H, I, J}.

[0117] In one embodiment, the probability threshold is set to 60%. That is, as long as (X∩Y) / (X∪Y)≥60% and (X∩Z) / (X∪Z)≥60%, by calculating (X∩Y) / (X∪Y)=7 / 9=77.78%>60% and (X∩Z) / (X∪Z)=7 / 10=70%>60%, then Y and Z can both be considered as peers of X.

[0118] Specifically, the steps for determining the clustering evaluation result of the clustered image set based on the first motion trajectory corresponding to the target object and the target object's peer objects in step S4 are as follows.

[0119] Please see Figure 7 , Figure 7 yes Figure 1 A flowchart illustrating a specific embodiment of step S4 in the provided data collection and evaluation method.

[0120] S41: Based on the first motion trajectory of the target object and the first motion trajectories corresponding to all objects in the same row as the target object, determine the second motion trajectory that is common to both the target object and the objects in the same row.

[0121] Specifically, the position corresponding to the second motion trajectory is the union of all positions contained in the first motion trajectories corresponding to the target object and the object in the same row, respectively.

[0122] Please see Figure 8 , Figure 8 This is a schematic diagram of the second motion trajectory corresponding to the target object X and the objects Y and Z in the same row.

[0123] In one embodiment, the positions of the second motion trajectory corresponding to X, Y, and Z are determined based on the positions corresponding to the target object X and the objects in the same row Y and Z. That is, X∪Y∪Z={A, B, C, C', D, E, F, G, H, I, J}.

[0124] In a preferred embodiment, in order to further purify the second motion trajectory, the positions in the second motion trajectory are examined, and specific step S4 further includes step S42.

[0125] S42: Denoise the corresponding points in the second motion trajectory based on the clustered images acquired from adjacent points in the first motion trajectory of the target object and / or peer objects.

[0126] Specifically, the corresponding points in the second motion trajectory are denoised through the following steps.

[0127] Please see Figure 9 and Figure 10 , Figure 9 yes Figure 7A flowchart illustrating a specific embodiment of step S42 in the provided data aggregation and evaluation method; Figure 10 This is a schematic diagram of the second motion trajectory after noise reduction processing.

[0128] S421: Traverse all points in the second motion trajectory and select one point as a candidate point.

[0129] Specifically, each point in the second motion trajectory is sequentially used as a candidate point.

[0130] S422: Determine whether the candidate site exists only in a first motion trajectory.

[0131] Specifically, the first motion trajectory of the candidate site is counted. The number of first motion trajectories of the candidate site determines whether further verification of the candidate site is needed.

[0132] If the candidate site exists in only one first motion trajectory, proceed directly to step S423; if the candidate site exists in at least two first motion trajectories, proceed directly to step S421.

[0133] S423: Determine the motion velocity corresponding to the candidate site based on the spatiotemporal information of the aggregated image acquired from the candidate site in the first motion trajectory and the spatiotemporal information of the aggregated image acquired from the previous point of the candidate site.

[0134] Specifically, if a candidate site exists only in one first motion trajectory, the motion velocity corresponding to the candidate site is determined based on the spatiotemporal information of the aggregated image captured at the candidate site within the first motion trajectory and the spatiotemporal information of the aggregated image captured at the adjacent preceding point of the candidate site. For example, if the selected candidate site is C', the motion velocity corresponding to the candidate site C' is determined based on the spatiotemporal information of the aggregated image captured at site C in the first motion trajectory of the same object Z and the spatiotemporal information of the aggregated image captured at site C'. The spatiotemporal information includes the capture time and the capture location.

[0135] In one specific embodiment, the time difference is determined based on the capture time of the clustered image acquired at site C and the capture time of the clustered image acquired at site C'; the distance difference is determined based on the coordinates of the capture site of the clustered image acquired at site C and the coordinates of the capture site of the clustered image acquired at site C'.

[0136] For example, given that the coordinates of point C are (a1, b1) and the coordinates of C' are (a2, b2), where a1 and a2 are longitudes and b1 and b2 are latitudes; and the Earth's radius is R = 6371.0 km, then the distance between the two points is: d = R * 3.1415926 / 180 * arcos(cos(b1)*cos(b2)*cos(a1-a2)+sin(b1)*sin(b2))*1000.

[0137] The motion velocity corresponding to candidate site C' is determined by the ratio of the distance difference to the time difference between the composite image acquired at site C and the composite image acquired at site C' in the first motion trajectory of the same object Z.

[0138] S424: If the movement speed of a candidate site is greater than the speed threshold, the candidate site is deleted.

[0139] Specifically, the speed threshold varies depending on the mode of transportation. For example, modes of transportation include walking, cycling, and driving. When the mode of transportation is walking, the speed threshold is the first threshold; when the mode of transportation is cycling, the speed threshold is the second threshold; and when the mode of transportation is driving, the speed threshold is the third threshold. The third threshold > the second threshold > the first threshold.

[0140] If the movement speed corresponding to candidate site C' is greater than the speed threshold, then candidate site C' is determined to be a noise site and is deleted.

[0141] By verifying each point in the second motion trajectory through the above steps, the updated second motion trajectory includes the points A, B, C, D, E, F, G, H, I, and J.

[0142] S43: Based on the second motion trajectory, the first motion trajectory of the target object, and the first motion trajectories corresponding to all objects in the same row as the target object, determine the clustering evaluation result of the clustered image set.

[0143] Specifically, the cluster evaluation results include cluster accuracy and cluster recall.

[0144] In one embodiment, the clustering evaluation results include clustering accuracy.

[0145] Please see Figure 11 , Figure 11 yes Figure 7 A flowchart illustrating a specific embodiment of step S43 in the provided data collection and evaluation method.

[0146] S4311: Based on the second motion trajectory obtained through denoising, determine the number of clustered images collected from the noise sites corresponding to the target object and the objects in the same row.

[0147] Specifically, the images acquired at each noise site can be multiple or a single image. In this embodiment, each noise site acquires only one image. The second motion trajectory in the denoising process includes sites A, B, C, D, E, F, G, H, I, and J, while the only noise site is C'. That is, the number of images acquired at the noise sites corresponding to the target object X and the same-row object Y is 0, while the number of images acquired at the noise sites corresponding to the same-row object Y is 1.

[0148] S4312: Determine the clustering accuracy of the clustered image set based on the number of clustered images collected from the noise sites corresponding to the target object and the objects in the same row.

[0149] Specifically, the number of clustered images acquired from the noise sites corresponding to the target object and all its peers is summed to obtain the number of noise images. The clustering accuracy of the image set is determined based on the ratio of the number of noise images to the total number of clustered images corresponding to the target object and all its peers. That is, clustering accuracy = 1 - (number of noise images / total number) * 100%.

[0150] In one embodiment, if there are three people traveling together, and the number of clustered images collected for each person corresponding to the noise site is 2, 0, and 1 respectively, and the number of clustered images for each person is 70, 85, and 80 respectively, then: Accuracy = 1 - (2 + 0 + 1) / (70 + 85 + 80) * 100% = 98.75%.

[0151] In another embodiment, the cluster evaluation results include cluster recall rate.

[0152] Please see Figure 12 , Figure 11 yes Figure 7 A flowchart illustrating another specific embodiment of step S43 in the provided data collection and evaluation method.

[0153] S4321: In response to a point in the second motion trajectory being a missed image location of the target object, determine whether the target object has an uncollected image at the missed image location based on the archived image of the target object's peer at the missed image location.

[0154] Specifically, a clustered image of a subject in the same field is selected at a missed capture site; based on the capture time of the clustered image of the selected subject in the same field at the missed capture site, images captured at the missed capture site within a preset time period are selected as candidate images; based on the similarity between each candidate image and an image containing the target object, it is determined whether the candidate image is an unclustered image of the target object captured at the missed capture site.

[0155] In one specific embodiment, if the similarity between a candidate image and an image containing the target object exceeds a similarity threshold, the candidate image corresponding to the similarity is retained; if the similarity between a candidate image and an image containing the target object does not exceed the similarity threshold, the candidate image corresponding to the similarity is discarded.

[0156] If no candidate image is retained for a missed image site, then no unaggregated image exists for that missed image site.

[0157] If at least two candidate images are retained corresponding to the missed image site, the candidate image with the highest similarity is selected as the uncollected image of the target object at the missed image site.

[0158] In one embodiment, Y and Z are objects traveling in the same direction as target object X. Based on the second motion trajectory and the first motion trajectory of target object X, it is determined that the aggregated image set of target object X does not include aggregated images containing the target object that were not captured at site B. Then, based on the capture time of the image containing target Z captured at site B by object Z traveling in the same direction at site B in the second motion trajectory, a preset time period is obtained by adding a preset duration forward and backward from this time as the center point. For example, the preset duration can be 30 seconds. If the capture time of the image containing target Z captured at site B is 2022-07-26 13:17:15, then all images captured at site B within the time period from 2022-07-26 13:16:45 to 2022-07-26 13:17:45 are selected. The similarity of each image within this time period is calculated with the image containing the target object. If the similarity exceeds a similarity threshold, the image is retained. All retained images are considered as non-aggregated images of target object X captured at site B. Alternatively, only the image with the highest similarity among the retained images can be selected as the uncollected image of the target object X at site B.

[0159] S4322: Traverse the missed images of the target object and the objects in the same row, and determine the uncollected images corresponding to the target object and the objects in the same row respectively.

[0160] By traversing the missed images of the target object and the objects in the same row using the method in step S4321 above, the uncollected images corresponding to the target object and the objects in the same row are determined respectively.

[0161] S4323: Determine the clustering recall rate of the clustered image set based on the number of missed sites corresponding to all non-clustered images and the number of sites corresponding to all clustered images, respectively, for the target object and the peer object.

[0162] Specifically, the first number of points is determined by summing the number of points corresponding to all clustered images of the target object and the number of points corresponding to all clustered images of all peer objects of the target object; the second number of points is determined by summing the number of points corresponding to all clustered images of the target object, the number of missed points corresponding to non-clustered images, and the number of points corresponding to all clustered images of all peer objects of the target object and the number of missed points corresponding to non-clustered images; and the clustered recall rate of the clustered image set is determined based on the ratio of the first number of points to the second number of points.

[0163] For example, if there are three people traveling together, and the number of loci corresponding to the clustered images for each person is 45, 48, and 39 respectively, then according to step 4322, the number of loci corresponding to the non-clustered and clustered images for each person is 58, 59, and 56 respectively. Therefore, the clustered recall rate is (45+48+39) / (58+59+56)*100% = 76.3%.

[0164] In this embodiment, based on the on-site production environment, a group of people traveling together with a high number of points within a certain time period is identified using existing data. The complete locations traversed by this group are obtained by taking the union of their movement trajectories. Then, by finding missing points on the trajectories, images that did not enter the main archive are retrieved. Finally, the archive aggregation effect (aggregation accuracy and aggregation recall) is statistically analyzed. It is worth mentioning that this series of operations can be automated using algorithms, eliminating the need for manual operation. The statistical analysis is based on data from the on-site production environment, taking into account the impact of image quality and site quality, resulting in more realistic and effective statistical results.

[0165] The image aggregation evaluation method provided in this embodiment acquires an image aggregation set, which consists of multiple images of a target object captured in different scenes. Each image aggregation set contains spatiotemporal information. Based on the spatiotemporal information of each image aggregation set, a first motion trajectory of the target object is determined. The first motion trajectory of the target object is compared with the first motion trajectory of another target to determine whether the target is a peer of the target object. The first motion trajectory includes at least two positions. In response to the target being a peer of the target object, the image aggregation evaluation result of the image aggregation set is determined based on the first motion trajectories corresponding to the target object and its peers. This application, by determining the peers of the target object and determining the image aggregation evaluation result of the image aggregation set based on the first motion trajectory of the target object and the first motion trajectories of all peers of the target object, can more comprehensively and accurately evaluate the image aggregation effect of the image aggregation set.

[0166] Please see Figure 13 , Figure 13This is a schematic diagram of the framework of an embodiment of the document aggregation evaluation device provided by the present invention. This embodiment provides a document aggregation evaluation device 60, which includes an acquisition module 61, a processing module 62, a comparison module 63, and an analysis module 64.

[0167] The acquisition module 61 is used to acquire a clustered image set, which consists of multiple clustered images of the target object in different scenes; each clustered image has spatiotemporal information.

[0168] The processing module 62 is used to determine the first motion trajectory of the target object based on the spatiotemporal information of each image.

[0169] The comparison module 63 is used to compare the first motion trajectory of the target object with the first motion trajectory of another target to determine whether the target is a peer of the target object; the first motion trajectory includes at least two positions.

[0170] The analysis module 64 is used to determine the clustering evaluation result of the clustered image set based on the first motion trajectory corresponding to the target object and the target object's peer object, in response to the target object being a peer object.

[0171] The file aggregation evaluation device provided in this embodiment determines the file aggregation evaluation result of the file aggregation image set by identifying the peer objects of the target object and the first motion trajectory of the target object and the first motion trajectory of all peer objects of the target object. This allows for a more comprehensive and accurate evaluation of the file aggregation effect of the file aggregation image set.

[0172] Please see Figure 14 , Figure 14 This is a schematic diagram of a terminal embodiment provided by the present invention. The terminal 80 includes a memory 81 and a processor 82 coupled to each other. The processor 82 is used to execute program instructions stored in the memory 81 to implement the steps of any of the above-described file evaluation method embodiments. In a specific implementation scenario, the terminal 80 may include, but is not limited to, a microcomputer or a server. In addition, the terminal 80 may also include mobile devices such as laptops and tablets, which are not limited here.

[0173] Specifically, processor 82 controls itself and memory 81 to implement the steps of any of the above-described aggregate evaluation method embodiments. Processor 82 may also be referred to as a CPU (Central Processing Unit). Processor 82 may be an integrated circuit chip with signal processing capabilities. Processor 82 may also be a general-purpose processor, digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component. A general-purpose processor may be a microprocessor or any conventional processor. Furthermore, processor 82 may be implemented using integrated circuit chips.

[0174] Please see Figure 15 , Figure 15 This is a schematic diagram of a framework of an embodiment of a computer-readable storage medium provided by the present invention. The computer-readable storage medium 90 stores program instructions 901 that can be executed by a processor. The program instructions 901 are used to implement the steps of any of the above-described embodiments of the file evaluation method.

[0175] In some embodiments, the functions or modules of the apparatus provided in this disclosure can be used to perform the methods described in the above method embodiments. The specific implementation can be referred to the description of the above method embodiments, and for the sake of brevity, it will not be repeated here.

[0176] The description of the various embodiments above tends to emphasize the differences between the various embodiments. The similarities or similarities between them can be referred to, and for the sake of brevity, they will not be repeated here.

[0177] In the several embodiments provided in this application, it should be understood that the disclosed methods and apparatus can be implemented in other ways. For example, the apparatus implementations described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.

[0178] 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.

[0179] 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.) or processor to execute all or part of the steps of the methods of 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.

[0180] The above are merely embodiments of the present invention and do not limit the scope of patent protection of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for evaluating data aggregation, characterized in that, include: Acquire a set of archived images, which consists of multiple archived images of the target object captured in different scenes; Each of the aforementioned aggregated images possesses spatiotemporal information; Based on the spatiotemporal information of each of the aggregated images, the first motion trajectory of the target object is determined; The first motion trajectory of the target object is compared with the first motion trajectory of another target to determine whether the target is a peer of the target object; the first motion trajectory includes at least two positions; In response to the target being a peer object of the target object, the clustering evaluation result of the clustered image set is determined based on the first motion trajectory corresponding to the target object and the peer object of the target object respectively; The step of determining the clustering evaluation result of the clustered image set based on the first motion trajectory corresponding to the target object and the target object's peer objects includes: Based on the first motion trajectory of the target object and the first motion trajectories corresponding to all the objects in the same row as the target object, a second motion trajectory corresponding to both the target object and the objects in the same row is determined; the position corresponding to the second motion trajectory is the union of all the positions contained in the first motion trajectories corresponding to both the target object and the objects in the same row. Based on the second motion trajectory, the first motion trajectory of the target object, and the first motion trajectories corresponding to all the peer objects of the target object, the aggregation evaluation result of the aggregation image set is determined.

2. The cluster assessment method according to claim 1, characterized in that, The first motion trajectory includes a starting point and an ending point. The step of comparing the first motion trajectory of the target object with the first motion trajectory of another target to determine whether the target is a peer of the target object includes: The start point and end point of the target object are compared with the start point and end point of the target. In response to the fact that the starting point and the ending point of the target object and the target are the same, it is determined whether the target is a peer object of the target object based on the points contained in the first movement trajectory of the target object and the points contained in the first movement trajectory of the target.

3. The cluster assessment method according to claim 2, characterized in that, Determining whether the target is a peer object of the target object based on the positions contained in the first motion trajectory of the target object and the positions contained in the first motion trajectory of the target object includes: The probability value of being in the same lane between the target object and the target is determined based on the ratio of the number of intersection points to the number of union points between the points corresponding to the first movement trajectory of the target object and the points corresponding to the first movement trajectory of the target object; the number of intersection points is the number of points corresponding to the intersection of the points in the first movement trajectory of the target object and the points in the first movement trajectory of the target; the number of union points is the number of points corresponding to the union of the points in the first movement trajectory of the target object and the points in the first movement trajectory of the target. If the probability value of the peer exceeds the probability threshold, then the target is determined to be a peer of the target object.

4. The cluster assessment method according to claim 1, characterized in that, Before the step of determining the clustering evaluation result of the clustered image set based on the second motion trajectory, the first motion trajectory of the target object, and the first motion trajectories corresponding to all the peer objects of the target object, the method further includes: The corresponding points in the second motion trajectory are denoised based on the clustered images acquired from the adjacent points in the first motion trajectory of the target object and / or the peer object.

5. The clustering evaluation method according to claim 4, characterized in that, The denoising process performed on the corresponding points in the second motion trajectory using the clustered image acquired based on the adjacent points in the first motion trajectory of the target object and / or the peer object includes: Traverse all the points in the second motion trajectory and select one of the points as a candidate point; Determine whether the candidate site exists in only one of the first motion trajectories; If the candidate site exists only in one of the first motion trajectories, the motion speed corresponding to the candidate site is determined based on the spatiotemporal information of the aggregated image collected from the candidate site in the first motion trajectory and the spatiotemporal information of the aggregated image collected from the previous candidate site. If the movement speed of the candidate site is greater than a speed threshold, the candidate site is deleted.

6. The cluster assessment method according to claim 4, characterized in that, The clustering evaluation results include the clustering accuracy rate; The process of determining the clustering evaluation result of the clustered image set based on the second motion trajectory, the first motion trajectory of the target object, and the first motion trajectories corresponding to all the peer objects of the target object includes: Based on the second motion trajectory obtained through denoising, determine the number of clustered images collected from the noise sites corresponding to the target object and the adjacent object, respectively. The clustering accuracy of the clustered image set is determined based on the number of clustered images collected from the noise sites corresponding to the target object and the peer object, respectively.

7. The cluster assessment method according to claim 6, characterized in that, Determining the clustering accuracy of the clustered image set based on the number of clustered images acquired from the noise sites corresponding to the target object and the peer objects respectively includes: The number of clustered images collected from the noise sites corresponding to the target object and all its peer objects in the same row is summed to obtain the number of noise images. The aggregation accuracy of the aggregation image set is determined based on the ratio of the number of noisy images to the total number of aggregation images corresponding to the target object and all the peer objects corresponding to the target object.

8. The cluster assessment method according to claim 1 or 4, characterized in that, The cluster evaluation results include the cluster recall rate; The process of determining the clustering evaluation result of the clustered image set based on the second motion trajectory, the first motion trajectory of the target object, and the first motion trajectories corresponding to all the peer objects of the target object includes: In response to a location in the second motion trajectory being a missed image location of the target object, it is determined whether the target object has an uncollected image at the missed image location based on the aggregated image acquired by the target object's peer at the missed image location; Traverse the missed images of the target object and the objects in the same row to determine the uncollected images corresponding to the target object and the objects in the same row, respectively; The clustering recall rate of the clustered image set is determined based on the number of missed sites corresponding to all the unclustered images corresponding to the target object and the peer object, respectively, and the number of sites corresponding to all the clustered images.

9. The cluster assessment method according to claim 8, characterized in that, The spatiotemporal information includes the capture time. The determination of whether the target object has an un-aggregated image at the missed image site based on the aggregated image of the target object's peers at the missed image site includes: Select one of the aforementioned peers to capture the archived image at the missed capture site; Based on the capture time of the archived images collected by the selected peer object at the missed capture site, images collected at the missed capture site within a preset time period are selected as candidate images. Based on the similarity between each candidate image and an image containing the target object, it is determined whether the candidate image is an uncollected image of the target object captured at the missed capture site.

10. The cluster assessment method according to claim 9, characterized in that, The step of determining whether a candidate image is an uncollected image of the target object captured at the missed capture site based on the similarity between each candidate image and an image containing the target object includes: If the similarity between the candidate image and the image containing the target object exceeds a similarity threshold, the candidate image corresponding to the similarity is retained. If at least two candidate images are retained corresponding to the missed image site, the candidate image with the highest similarity is selected as the uncollected image of the target object collected at the missed image site.

11. The clustering evaluation method according to claim 10, characterized in that, The step of determining the clustering recall rate of the clustered image set based on the number of missed sites corresponding to all the unclustered images corresponding to the target object and the peer object respectively, and the number of sites corresponding to all the clustered images, includes: The number of the first point is determined by summing the number of the sites corresponding to all the clustered images of the target object and the number of the sites corresponding to all the clustered images of all the same-row objects of the target object. The second number of sites is obtained by summing the number of sites corresponding to all the archived images of the target object, the number of missed sites corresponding to the unarchived images, and the number of sites corresponding to all the archived images of all the peer objects of the target object, and the number of missed sites corresponding to the unarchived images. The cluster recall rate of the clustered image set is determined based on the ratio of the number of the first site to the number of the second site.

12. A data collection and evaluation device, characterized in that, include: The acquisition module is used to acquire a set of archived images, which are multiple archived images of the target object collected in different scenes; Each of the aforementioned aggregated images possesses spatiotemporal information; The processing module is used to determine the first motion trajectory of the target object based on the spatiotemporal information of each of the aggregated images; The comparison module is used to compare the first motion trajectory of the target object with the first motion trajectory of another target to determine whether the target is a peer of the target object; the first motion trajectory includes at least two positions; The analysis module is configured to, in response to the target being a peer object of the target object, determine the clustering evaluation result of the clustered image set based on the first motion trajectories corresponding to the target object and the peer objects of the target object respectively; and is further configured to, based on the first motion trajectory of the target object and the first motion trajectories corresponding to all the peer objects of the target object respectively, determine a second motion trajectory jointly corresponding to the target object and the peer objects; the position corresponding to the second motion trajectory is the union of all positions contained in the first motion trajectories corresponding to the target object and the peer objects respectively; Based on the second motion trajectory, the first motion trajectory of the target object, and the first motion trajectories corresponding to all the peer objects of the target object, the aggregation evaluation result of the aggregation image set is determined.

13. A terminal, characterized in that, The terminal includes a memory, a processor, and a computer program stored in the memory and running on the processor, the processor being configured to execute program data to implement the steps in the file evaluation method as described in any one of claims 1 to 11.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the document evaluation method as described in any one of claims 1 to 11.

Images