Target object clustering method and device

Abstract

The invention relates to the technical field of image processing, in particular to a target object clustering method and device, and the target object clustering method comprises the steps: obtaininga monitoring video, intercepting sub-images comprising target object information from an image of the monitoring video, and recording the frame number of each sub-image; then, extracting a feature vector of each sub-image, and based on the extracted feature vector of each sub-image, dividing the sub-images comprising the same target object information into the same category set; and finally, determining a to-be-adjusted category set based on the frame numbers of the sub-images included in the divided category sets, and adjusting the sub-images included in the to-be-adjusted category set. Through the target object clustering method, the clustering accuracy of the target object can be improved.

Description

technical field [0001] The present application relates to the technical field of image processing, and in particular to a target object clustering method and device. [0002] This case is a divisional application of the patent application with application number 201811544296.8. Background technique [0003] In the application fields of video surveillance, security, unmanned driving, etc., it generally involves the detection of target objects in the surveillance video, such as pedestrians or vehicles appearing in the surveillance video. Specifically, in some specific application scenarios, for example, when determining the activities of the target object in the area monitored by the surveillance video, it is necessary to filter out all images with the target object from the surveillance video, which involves the monitoring of the surveillance video Cluster images with target objects in them. [0004] In the prior art, commonly used clustering algorithms are mainly K-means a...

AI Technical Summary

Problems solved by technology

[0004] In the prior art, commonly used clustering algorithms are mainly K-means and KD-tree, etc., but such clustering algorithms often first need to determine a category threshold in order to determine how many categories the target objects need to be clustered into, but When clustering the target objects in the surveillance video, it is often difficult to determine how many categories the target objects need to be clustered into. Therefore, the selection of this category threshold may have a large error, resulting in a low clustering accuracy