Target character distance measurement learning-oriented character re-recognition method

Abstract

The invention provides a target character distance measurement learning-oriented character re-recognition method, and relates to the technical field of video processing. According to the method, a non-target query character is imported onto a target task data set to serve as an auxiliary constraint, and under a distance measurement and transfer learning framework, distance measurement orienting tothe target query character is online learnt, so that the character re-recognition problem of comparing multiple frames of images is solved. According to the method, the influences caused to statistical model training and learning on target task data sets by diversity and difference of data sets when third-party character re-recognition task data sets are imported are avoided, the problem of statistical model overfitting caused by less training data is relieved, loss functions are defined at near distances of target character feature instances, the problem that an in-class variance of a same character feature instance is greater than an inter-class variance is overcome, and character re-recognition tasks for comparing multiple frames of images are solved.

Description

technical field [0001] The invention relates to the technical field of video processing, in particular to a person re-identification method. Background technique [0002] Person re-identification is to correlate and match the person images captured by different cameras, and it is used to understand the behavior of the target person in the camera monitoring network. It is an important technology in the camera monitoring system. With the widespread deployment of camera networks and video surveillance systems, person re-identification technology has received more and more attention. [0003] Due to the low resolution of the surveillance camera and the long distance of the surveillance camera, biological information such as the face and gait of the person cannot be used in the person re-identification technology. Therefore, the current research work assumes that the person's appearance information such as clothing does not change significantly in the camera monitoring network, ...

AI Technical Summary

Problems solved by technology

On the one hand, it is very difficult to obtain the data set of the person re-identification task discussed above. When the target task scale is large, it is still difficult to introduce a third-party data set to meet the needs of the target task; Due to the diversity and difference of data, it is difficult to fully apply the data of third-party datasets to the target tasks

At the same time, the above research methods focus on the person re-identification task of comparing a single image between the target person and the candidate person. Since the intra-class variance of multiple frames of images of the same person is greater than the inter-class variance of images, these research methods cannot be directly applied to more Challenging multi-frame image comparison task for person re-identification

Images