Detection System Of Potential Accidents Around Transmission Line Based On Image Difference

With the development of the national economy,all walks of life in the demand for electricity continues to expand,the loss due to power failure which is caused by sundry man-made and natural accidents is also increasing.How to find and prevent the natural disasters and accidents that endanger the transmission line,and to ensure the normal supply of electricity,has become the focus of attention.Threats to the safety of transmission lines mainly include ultra high operating vehicles,under line construction,forest fires,etc.Relevant departments need to constantly patrol the line to search for possible hidden dangers,then solve it.At present,the common route inspection methods include manual inspection,UAV(unmanned aerial vehicle)inspection,online video surveillance,etc.Manual inspection requires a lot of manpower and its efficiency very low.The flight distance of UAV is limited,so that the UAV inspection is difficult to popularize in remote unmanned area.Online video surveillance is a new way to patrol line,and there are some intelligent video analysis systems that can detect some hidden dangers.It has the characteristics of flexibility and high efficiency,but it needs to consume a lot of transmission traffic to take video back to the server,which restricts its development.Recently,a company has proposed an improved scheme for online video surveillance inspection,which is still monitored by installing cameras on transmission towers.Unlike traditional video shooting,in this scenario,the camera takes pictures to the server at intervals of time.And then the staff view the picture to determine whether there is any hidden dangers.This can not only obtain the scene information,but also greatly reduce the loss of equipment and consumption of electricity flow.The problem of the scheme is that there is no corresponding potential accidents detection system based on image analysis,and all the observation and judgment work is accomplished by manual.After the installation of more than 10,000 equipment,staff need to observe nearly 100,000 pictures every day,reflecting the problem of excessive workload.Based on the present situation and demand of the above-mentioned scheme,our team cooperated with the company to develop a detecting system of hidden dangers of transmission lines based on the analysis of image difference.The system utilizes the method of image processing and machine learning,analyzes the image of transmission line,finds the potential accidents and then alarms.This paper mainly describes the detection algorithm design and system realization of the potential accidents detecting system.The focus of the study is to further sift and identify potential accidents based on image differences.The details are as follows:(1)Classify the potential accidents according to their characteristics and detection methods.Through the observation and analysis of the transmission line environment and various hidden dangers,we divide the potential accidents into instant hidden dangers,such as tower Crane,crane,engineering vehicle and fire,and the long-term potential accidents which include under line construction,parking lot and stacking.And then we designed several different methods for the detection.(2)Aiming at the characteristics of various potential accidents,we design and implement the corresponding potential accidents detection algorithm.It mainly contains the method of detecting crane and crane in sky based on background modeling,the method of detecting fire by color and texture information,the method of identifying engineering vehicles using convolution neural network and the method to detect long-term potential accidents of construction,parking lot and stacking by using statistic model.(3)We built a potential accidents detection system based on the potential accidents detection algorithm.The system mainly realizes the function of detecting all kinds of potential accidents in pictures.This verifies the validity of the potential accidents detection algorithm.According to the needs of the actual situation,the system also has the functions of monitoring,alarm,manual confirmation,uploading and history inquiry.After using the system,the staff only need to observe the alarm pictures and confirm it.That greatly reduced the labor workload.At present,the inspection scheme using our detection system has been popularized in the 17 cities,and ten thousands of camera have been deployed.There is very little omission in the running process,and the alarm rate is only about 5% by statistics.It means that 95% of the labor workload is reduced,with greater practical significance.