Personnel recognition method based on multi-visual-feature fusion

Abstract

Provided is a personnel recognition method based on multi-visual-feature fusion. Firstly, a monitoring video is analyzed, information of related personnel is extracted, features are described, and finally searching is carried out to obtain a recognition result of the related personnel according to an input video image. The method comprises the three operation stages or steps of video tracking processing, human body block mass processing and personnel recognition. The method can well extract proper personnel images from the video automatically and carry out preprocessing; the existing problems of feature description are solved correspondingly, relatively general features are selected, the features are regrouped, and new features are fused. The method achieves innovation on the aspects of removing a background part, blocking the human body images and extracting semantic color features. The testing result of the monitoring video by multiple times of simulation examples shows that the method is easy to implement, convenient to use, effective and good in recognition effect, and therefore the method has good application and popularization prospects.

Description

technical field [0001] The invention relates to a person recognition method based on multi-visual feature fusion, and belongs to the technical fields of computer vision, digital image processing, multimedia information processing and video monitoring. Background technique [0002] Person recognition is a research hotspot in the field of computer vision. It uses technology based on biometrics to solve the problem of identifying people, thus giving birth to the branch field of person recognition based on multiple features such as face, iris, fingerprint and gait. . However, in most video surveillance scenarios, accurate biometrics are often not obtained. Moreover, the resolution and frame rate of the video are relatively low, and the background environment is complex, which greatly reduces the effect of person identification based on biometrics, or even cannot conduct. In this context, a new branch of person recognition: appearance-based recognition research was born. [00...

AI Technical Summary

Problems solved by technology

Moreover, if it is a person image obtained by a non-manual method, it often contains more background parts

At this time, even after preprocessing, the background part will still have adverse effects on subsequent feature extraction and feature comparison, and ultimately affect the recognition results

In the second step of processing, the selected features are often not universal, and are only suitable for use in certain specific environments, and the description of personnel information is not comprehensive enough, the difference between different personnel is not obvious enough, and the recognition results are relatively poor

[0006] In addition, the scope of existing person recognition work for surveillance video is very limited, and most of them are only for a single image

Images