| In recent years,with the maturer video surveillance processing technology,video surveillance system has begun to develop in the direction of intelligence.As a personal protective headgear,helmet is required to wear during the factory’s construction operations.However,currently to detect whether wearing helmets or not for workers mainly depends on manual inspection and the efficiency is very low.So if we can utilize intelligent video surveillance system to implement the detection of wearing helmets,and realize the alarm and linkage control for no wearing helmets condition,it is of great significance to the safety of the workers in the factory.In view of the above situation,some researchers at home and abroad have carried out research on the detection system of wearing helmets.However,most of the proposed methods have problems including high time complexity,poor robustness and low accuracy.In the complicated environment of the factory,it is difficult to achieve better results.Considering the above problems,based on the specific application scenarios and requirements of the system,this paper designs and implements a detection system of wearing helmets based on intelligent video surveillance.For the detection module of system,we exploit a method integrating multi-layer features of the inherent feature map pyramid of the deep convolution neural network and performing hierarchical prediction.The method is used to improve the Faster R-CNN.Finally,we design a deep-learning based detection algorithm of wearing helmets.In the personnel tracking module,we propose a tracking algorithm which combines depth convolution neural network information from the detection module and the kernel correlation filter tracking algorithm.In order to evaluate the system,this paper has conducted a large number of experiments.The results prove that our system is a real-time,accuracy,robustness and reliability system and can meet the actual needs. |
