CN-116152535-B - Target object clustering method and device, electronic equipment and readable storage medium

CN116152535BCN 116152535 BCN116152535 BCN 116152535BCN-116152535-B

Abstract

The invention provides a target object clustering method, a target object clustering device, electronic equipment and a readable storage medium, wherein the method comprises the steps of obtaining a plurality of target object files in a preset time period of a target area, wherein each target object file in the plurality of target object files comprises a plurality of snapshot devices for taking a snapshot of a corresponding target object, snapshot time and longitude and latitude information of each snapshot device in the plurality of snapshot devices, managing the corresponding snapshot devices according to the plurality of target object files to obtain a managed file, counting association relations between two snapshot devices in the plurality of snapshot devices according to the managed file, digging out snapshot data missing from the files to be tested in the managed file according to the association relations, and carrying out file aggregation on the corresponding file according to the snapshot data. The method is used for improving the accuracy and recall rate of target object clustering.

Inventors

WANG KAIYAO
CHEN LILI
ZHOU MINGWEI

Assignees

浙江大华技术股份有限公司

Dates

Publication Date: 20260512
Application Date: 20230220

Claims (8)

1. A method for clustering target objects, comprising: Acquiring a plurality of target object files in a preset time period of a target area, wherein each target object file in the plurality of target object files comprises a plurality of snapshot devices for snapshot of a corresponding target object, and snapshot time and longitude and latitude information of each snapshot device in the plurality of snapshot devices; treating the corresponding snapshot equipment according to the target object files to obtain treated files; according to the treated file, calculating the association relationship between two snapshot devices in the plurality of snapshot devices; According to the association relation, mining out snapshot data missing from the files to be tested in the treated files, and carrying out file gathering on the corresponding files according to the snapshot data; the method for controlling the snapshot equipment according to the target object files comprises the steps of: Counting a snapshot time difference sequence of each target object file in the plurality of target object files passing through a space between two snapshot devices, wherein the snapshot time difference sequence comprises a plurality of snapshot time differences, and each snapshot time difference is a time difference between snapshot times of the two snapshot devices for capturing the same target object; taking the median of the snapshot time difference sequence as the real passing time between the two snapshot devices; Comparing the real passing time with the preset passing time between the two snapshot devices based on map crawling to obtain a comparison result; and if the comparison result shows that the deviation between the real passing time and the preset passing time is larger than a preset time threshold, treating the two snapshot devices, and recording longitude and latitude information of the two snapshot devices again to obtain treated files.
2. The method of claim 1, wherein the counting the association between two snapshot devices of the plurality of snapshot devices according to the treated archive comprises: Counting the conversion times between two snapshot devices in each file in the treated file, and accumulating the conversion times between the two snapshot devices of all files in the treated file to obtain the total conversion times between the two snapshot devices; Determining the conversion rate between the two snapshot devices according to the conversion times and the total conversion times; If the conversion rate is greater than a preset conversion rate threshold, the association relationship between the two snapshot devices indicates that the two snapshot devices are continuous snapshot devices.
3. The method of claim 2, wherein the mining out the snapshot data missing from the archive to be tested in the managed archive according to the association relationship comprises: If the association relation indicates that the first snapshot device and the second snapshot device in the treated file are continuous snapshot devices, and the file to be tested in the treated file is snapshot by the first snapshot device at a first time and is snapshot by a third snapshot device at a second time after the first time, the file to be tested is necessarily snapshot by the second snapshot device at a third time between the first time and the second time; if the snapshot data of the file to be tested is missed at the third moment, a preset file gathering threshold is reduced, and file gathering is carried out on the snapshot data and the file to be tested.
4. The method of claim 1, wherein the counting the association between two snapshot devices of the plurality of snapshot devices according to the treated archive comprises: Counting the time difference of each file in the treated files passing through the fourth snapshot equipment and the fifth snapshot equipment; if the time difference is smaller than the preset passing time, indicating that the corresponding file effectively passes through the fourth snapshot equipment and the fifth snapshot equipment once; If the corresponding file passes through the sixth snapshot device during the period of passing through the fourth snapshot device and the fifth snapshot device, the association relationship between the fourth snapshot device and the fifth snapshot device indicates that the corresponding file passes through a unique preset path between the fourth snapshot device and the fifth snapshot device.
5. The method of claim 4, wherein the mining out the snapshot data missing from the archive to be tested in the managed archive according to the association relationship comprises: Counting the times of the treated archives passing through the sixth snapshot device during the period of effectively passing through the fourth snapshot device and the fifth snapshot device, and occupying the proportion of the total times of the effective passing through the fourth snapshot device and the fifth snapshot device; If the ratio is greater than a preset frequency threshold, indicating that the sixth snapshot device is needed to pass through during the process of passing through the fourth snapshot device and the fifth snapshot device; If the file to be detected in the treated file effectively passes through the fourth snapshot equipment and the fifth snapshot equipment, but no snapshot data of the sixth snapshot equipment exist in the period of effectively passing through the fourth snapshot equipment and the fifth snapshot equipment, the condition that the file to be detected is in missing file gathering is indicated; and reducing a preset gear gathering threshold, and carrying out gear gathering again on the snapshot data which is effectively subjected to the snapshot by the sixth snapshot equipment during the period of the fourth snapshot equipment and the fifth snapshot equipment and the file to be tested.
6. A target object clustering device, comprising: the acquisition unit is used for acquiring a plurality of target object files in a preset time period of a target area, wherein each target object file in the plurality of target object files comprises a plurality of snapshot devices for snapshot of a corresponding target object, and snapshot time and longitude and latitude information of each snapshot device in the plurality of snapshot devices; The control unit is used for controlling the corresponding snapshot equipment according to the target object files to obtain controlled files; The statistics unit is used for counting the association relation between two snapshot devices in the plurality of snapshot devices according to the treated files; The file gathering unit is used for excavating out snapshot data missing from the files to be tested in the treated files according to the association relation, and gathering files of the corresponding files according to the snapshot data; wherein, the treatment unit is specifically used for: Counting a snapshot time difference sequence of each target object file passing between two snapshot devices before and after the target object files in the target object files, wherein the snapshot time difference sequence comprises a plurality of snapshot time differences, and the snapshot time differences are time differences between the snapshot times of the two snapshot devices for capturing the same target object; taking the median of the snapshot time difference sequence as the real passing time between the two snapshot devices; Comparing the real passing time with the preset passing time between the two snapshot devices based on map crawling to obtain a comparison result; and if the comparison result shows that the deviation between the real passing time and the preset passing time is larger than a preset time threshold, treating the two snapshot devices, and recording longitude and latitude information of the two snapshot devices again to obtain treated files.
7. An electronic device comprising a memory and a processor, the memory having stored thereon a computer program executable on the processor, the computer program, when executed by the processor, implementing the method of any of claims 1-5.
8. A computer-readable storage medium, in which a computer program is stored, characterized in that the method according to any one of claims 1 to 5 is implemented when the computer program is executed by a processor.

Description

Target object clustering method and device, electronic equipment and readable storage medium Technical Field The present invention relates to the field of image processing technologies, and in particular, to a target object clustering method, a target object clustering device, an electronic device, and a readable storage medium. Background With the continuous progress of society, target object files have become an important means for targeting related departments. The existing target object clustering technology mainly comprises the steps of analyzing target object pictures which are captured by cameras in an area in a period of time through a deep learning technology to form feature vectors, performing similarity calculation based on the feature vectors, and performing target object file aggregation to form target object files corresponding to each target object. The target object can be locked according to the track information of the target object file. However, due to different snapshot distances, snapshot angles, snapshot devices and the like, the similarity of the pictures of the same target object may not reach a file gathering condition, for example, the similarity of the pictures of the target object corresponding to the same target object may be slightly lower than a clustering threshold, so that clustering of the pictures of the target object is omitted, and if the clustering of the pictures of the target object corresponding to all the target objects is reduced according to a uniform standard, a clustering error of the pictures of the target object may be caused. Therefore, the accuracy and recall rate of target object focus are low. Disclosure of Invention The invention provides a target object clustering method, a target object clustering device, electronic equipment and a readable storage medium, which are used for improving the accuracy and recall rate of target object clustering. In a first aspect, an embodiment of the present invention provides a target object clustering method, including: Acquiring a plurality of target object files in a preset time period of a target area, wherein each target object file in the plurality of target object files comprises a plurality of snapshot devices for snapshot of a corresponding target object, and snapshot time and longitude and latitude information of each snapshot device in the plurality of snapshot devices; treating the corresponding snapshot equipment according to the target object files to obtain treated files; according to the treated file, calculating the association relationship between two snapshot devices in the plurality of snapshot devices; And according to the association relation, mining out the snapshot data missing from the files to be tested in the treated files, and carrying out file aggregation on the corresponding files according to the snapshot data. In one possible implementation manner, the managing the corresponding snapshot device according to the plurality of target object files to obtain a managed file includes: Counting a snapshot time difference sequence of each target object file in the plurality of target object files passing through a space between two snapshot devices, wherein the snapshot time difference sequence comprises a plurality of snapshot time differences, and each snapshot time difference is a time difference between snapshot times of the two snapshot devices for capturing the same target object; taking the median of the snapshot time difference sequence as the real passing time between the two snapshot devices; Comparing the real passing time with the preset passing time between the two snapshot devices based on map crawling to obtain a comparison result; and if the comparison result shows that the deviation between the real passing time and the preset passing time is larger than a preset time threshold, treating the two snapshot devices, and recording longitude and latitude information of the two snapshot devices again to obtain treated files. In one possible implementation manner, the counting, according to the treated file, an association relationship between two snapshot devices in the plurality of snapshot devices includes: Counting the conversion times between two snapshot devices in each file in the treated file, and accumulating the conversion times between the two snapshot devices of all files in the treated file to obtain the total conversion times between the two snapshot devices; Determining the conversion rate between the two snapshot devices according to the conversion times and the total conversion times; If the conversion rate is greater than a preset conversion rate threshold, the association relationship between the two snapshot devices indicates that the two snapshot devices are continuous snapshot devices. In one possible implementation manner, the mining the snapshot data missing from the archive to be tested in the managed archive according to the association relationship includes: If