Research On Recognition Algorithm Of Safety Helmet Wear Of Substation Personnel

With the development of machine vision,image processing,pattern recognition and other related fields,video image recognition technology has been widely used in various fields of social life.For example,in the electric power production system,a lot of power research scholars put forward the application of video image recognition technology to Substation Video Monitoring System in order to realize the intelligent substation.As is known to all,the safety helmet is an important safety protection tool for substation workers.In the process of electric power construction,electric power workers are required to wear.However,in recent years,the substation safety incidents occur frequently,the fundamental reason lies in the power of workers operating in violation of power safety norms,which is not according to the standard wear safety helmet is one of the important reasons.In order to prevent the occurrence of electric power safety incidents,it is more and more urgent to develop a set of intelligent video surveillance system which can automatically identify abnormal situation and timely alarm,such as staff not wearing a safety helmet.In this paper,based on the background of some parts of the power grid substation,this paper makes an in-depth study on the identification algorithm of the helmet wearing status of the power workers in the workplace.The main difficulties of this paper are as follows:(1)the substation is usually in the outdoor environment.(2)the substation power equipment is more and more complex,which is not conducive to the detection of safety helmet.(3)the network camera is far away from the substation area,the video image resolution is low.(4)the helmet color and the background color of substation scene are lower.In order to solve the above problems,this paper first introduces the commonly used algorithms in the process of the helmet wearing state recognition and collects video samples from the actual substation scene.In this paper,the advantages and disadvantages of the commonly used algorithms in the helmet wearing state recognition are compared in detail.We compare the advantages and disadvantages of the commonly used algorithms in the helmet wear state recognition.In this paper,we use a helmet wear state recognition algorithm.Firstly,a moving object detection algorithm based on pixel level visual background extraction(ViBe)is used.The advantage of the algorithm: the speed of background modeling fast,strong anti-interference ability,updating the background model with stochastic update strategy,which can effectively overcome the substation video image pixel variations.Then,In view of the extracted foreground target,a new method based on the aspect ratio of human body is used to locate the helmet wearing area.Finally,an identification algorithm based on deformable component model(DPM)is used to detect the wearing state of the helmet.The basic idea of this algorithm,which uses the sliding window mechanism and the idea of the image Pyramid,from the multi-scale,multi component,multi angle constructs feature model and uses the implicit support vector machine training helmet wearing state recognition model,which matches the target to determine whether the staff wear a helmet.The experimental results show that the detection algorithm proposed in this paper can meet the requirements of the performance of the helmet wearing state recognition.The algorithm provides a new idea for the application of helmet wearing state recognition technology in substation intelligent video surveillance system.