Design And Implementation Of Abnormal Behavior Detection System For Power Operation

In the production and construction of electric power,electric accidents will inevitably occur,which will not only cause economic losses,but also cause personal injury.After investigation,it was found that the main cause of power accidents was illegal operation.At present,the power job site mainly adopts manual supervision to correct the illegal operations of workers.The manual supervision method has disadvantages such as low efficiency and strong subjectivity.In order to make up for the lack of manual supervision,we have designed and developed a deep learning-based abnormal behavior recognition system for electric power operations to supervise workers on the job site in real time.When workers are found to have behaviors that do not comply with the electrician's safety operating procedures,they are given voice reminders.In recent years,deep learning has developed rapidly and has performed very well in areas such as target detection.This article uses deep learning methods combined with hardware devices to produce automatic identification equipment for abnormal behaviors in power scenarios.The specific tasks are as follows:(1)Aiming at power operation scenarios,a target detection method based on dual-stream Faster R-CNN is proposed.It realizes the target detection function at the electric work site,mainly detecting whether electric workers are wearing safety protection objects that must be worn during work in accordance with the requirements of electrician safety operating procedures.If they are not worn correctly,they will be exposed to abnormal behaviors.A voice warning is issued.(2)Aiming at the dangerous area of the electric power operation site,a method of using ultra-wideband to provide sensing signals and combining machine learning for classification and positioning is proposed,which realizes the function of reminding the dangerous area.Since there is a high-voltage dangerous environment at the electric power operation site,in order to prevent workers from entering the dangerous area and being personally injured,the function of reminding the dangerous area will warn the workers when they enter the dangerous area.(3)Aiming at the blind spot of the camera,a pedestrian tracking algorithm based on Raspberry Pi and robotic arm is proposed,which realizes the function of real-time target tracking for workers.The target tracking function is mainly to solve the problem of the blind spot of the camera.This function can adjust the camera's angle of view to follow the worker's movement when the electric worker is moving,so as to ensure that the image information of the worker's on-site operation can always be obtained.