Research and application of security video labeling platform

Abstract

With the rapid development of science and technology, the intelligent security technology has developed rapidly. As far as the deep learning method used in the field of intelligent security is concerned, image marking is a heavy and complicated work. The current security video is usually manually completed, which increases the operation and maintenance cost of security video system. For this reason, this paper proposes a semi-automatic annotation method based on neighborhood rough set for the complexity of distorted image annotation. The division of theoretical domain and the definition of correlation degree are redone. The aim is to start with the manual annotation of the data source, improve the quality of the annotation, and solve the difficulty of the annotation caused by the standard deviation of the subjective evaluation of the annotator and the blurring of the boundary between the annotated words. The application example shows that the security video labeling platform can realize semiautomatic video labeling and significantly improve the accuracy rate of security video labeling.