Application And Research Of Image Recognition Technology In The Monitoring System

In this paper, a preliminary study of helmets and body positioning in helmets recognition system has been done. First, we take measures based on HOG and SVM to position the body. Secondly, the aim of helmet positioning is achieved on the basis of Haar-like features and AdaBoost sorter. Lastly, according to the positioning of body and helmets, we can know whether workers on construction site wear helmets or not. As the real-time performance is unsatisfying when testing on body by using HOG, this paper improves the method of feature extraction. Experimental results show that the improved human detection algorithm has good real-time performance. In general, although the system detects both of the human body and the helmet, it cannot be assured operation workers are wearing helmets. Therefore, it is necessary to use spatial information of the two for a more accurate judgment. Experiments show that the system basically meets the need of monitoring operating workers at risk situations to wear safety helmets.