A people image search method based on deep reinforcement learning

Abstract

The invention provides a people image search method based on deep reinforcement learning. The method comprises the steps of: S1, defining various kinds of action for adjustment of a search area of a target image; S2, building a configurable deep model; S3, collecting training samples and training a strategy selection network and a value network by using the training samples; S4, inputting a reference images and to the-be-tested target image into the deep model and initializing the search area of target image as the whole image; S5, extracting features of the reference image via a feature extracting network; S6, extracting features in the search area of the target image via the feature extracting network, and fusing the same and the features of the reference image to form fusion features. The method treats pedestrian detection and people reidentification as the same task and only needs multiple times of action selection to judge whether a target person is found without an extra candidate frame, thereby having high time efficiency.

Description

technical field [0001] The invention relates to the field of computer vision recognition, in particular to a person image search method based on deep reinforcement learning. Background technique [0002] Video surveillance is an important way to improve public safety management. With the maturity of image acquisition technology and the decline of storage technology costs, more and more camera networks are deployed in public places, such as airports, railway stations, shopping malls, and university campuses. As a result, a large number of video resources have been generated. Using manual screening and processing is not only inefficient and consumes a lot of manpower and material resources, but also may introduce some human factors, resulting in some deviations. Human beings are one of the most important targets in video surveillance, and related topics around human beings have received extensive attention in the field of computer vision. [0003] The main task of person imag...

AI Technical Summary

Problems solved by technology

[0006] 2) The mainstream pedestrian detection algorithm needs to provide a large number of candidate frames, and there is a large amount of computational redundancy

Images