Anti-Tailing AB-Door Detection Based on Motion Template Algorithm
Yuwen Jia, Жипенг Ли · 2019
The product of anti-tailing AB-Door is a kind of high-tech product among the security entrance guard systems, which is widely demanded in the high-end application fields of finance (banks), security (prisons), official business, general places(factories) where anti-submerged crime exists. The main task for the system to function well is the accuracy of human's detection. This paper introduces an effective system to detect multi-person between AB-door based on Motion History Image (MHI) and motion segmentation. The direction of movement is then calculated depends on MHI which is essential for the following operation for the doors. In order to solve the problem that people cannot be detected when no movement occurs, we adopt novel original Token and Separation Degree Analysis algorithm to optimize our method. Because of the complexity of natural scene, a perfect Intelligent Video System tends to be based on different algorithm policies to deal with a variety of situations unpredicted beforehand. We tried to conquer the statistics of person number together with the disturbance of half-transparent glasses, free poses, light changing and imaging distortion in a tiny fixed area between two doors, which were supposed to contain at most single. We measured the performance of the algorithm in the above scenarios which demonstrated high accuracy and strong adaptability.