Video sequence encoding and decoding method for video pedestrian re-identification

Abstract

The invention discloses a video sequence coding and decoding method for video pedestrian re-identification, and the method comprises the following steps: in a training stage, fusing label picture features and video features, inputting the fused features into a generator, then employing a label picture as a reconstruction label, and restraining a key frame generated by the generator through image reconstruction loss; and then sending the generated key frame to an image feature extraction module for video feature recovery, and constraining the recovered video features through feature reconstruction loss, so that the video features are consistent with original video features in performance. In an application stage, K frames of pictures are selected by using an HSVTopK method to be used for generating key frames, and then the generated key frames are stored in equipment so as to reduce the storage overhead. When retrieval is needed, the image feature extraction module is used for carryingout video feature recovery on the generated key frame, and the recovered features reserve the performance of the video features and are used for retrieval matching of pedestrians.

Description

technical field [0001] The invention belongs to the field of computer vision image retrieval, in particular to a video sequence encoding and decoding method for video pedestrian re-identification. Background technique [0002] Pedestrian re-identification is aimed at retrieving pedestrians specified by users from a series of cross-camera surveillance videos; it is widely used in smart cities and security monitoring. [0003] According to the number of different input pictures, person re-identification can be divided into video person re-identification and image person re-identification. Compared with image person re-identification that uses a single frame image as input, video person re-identification uses video sequences as input and is more robust to environmental disturbances. However, video person re-identification needs to store a large number of video sequences, which will lead to huge storage overhead in practical applications and increase the application cost of vid...

AI Technical Summary

Problems solved by technology

However, video person re-identification needs to store a large number of video sequences, which will lead to huge storage overhead in practical applications and increase the application cost of video person re-identification.

At the same time, due to the different lengths of each video sequence in the application stage, it is not suitable for batch processing, resulting in a large computational overhead

Images