Research On Transmission Line Recognition And Tower Detection Technology Based On Image Processing

With the rapid development of the national economy,the demand for electric energy is increasing,and high-voltage transmission lines are increasing day by day.Due to long-term exposure to the harsh natural environment and natural disasters such as strong winds,showers and snowstorms,transmission lines are prone to breakdowns.If they are not found and maintained in time,power outages will be caused,which will directly affect industrial production and people’s daily lives,resulting in direct and indirect economic losses.At the present stage,manual inspection is the main method of transmission line inspection,which has high cost,large workload and low efficiency.In view of the above problems,the combination of unmanned aerial vehicle and image processing technology brings ideas to electric power inspection to solve the above problems.For this reason,this paper mainly studies the methods of transmission line recognition and tower detection,and the main work and achievements include the following three aspects.(1)According to the image characteristics of transmission line captured by aerial photograph of unmanned aerial vehicle,the image preprocessing method is studied.Through the experimental comparison,the weighted average method is selected to gray the image,the Gaussian filter is used to smooth the image,and the histogram equalization is used to enhance the image.In order to solve the problem of background noise in edge image,the connected region of edge image is analyzed to remove the background noise and weaken the interference of background noise to the result of transmission line recognition.(2)The real-time recognition method of transmission line is studied.The standard hough transform is easy to generate a plurality of parallel or intersecting straight lines in the recognition of transmission lines.An improved hough transform algorithm is proposed to make the recognition of transmission lines meet the single straight line response criterion,and the transmission line information can be accurately and completely extracted.Aiming at the problem that the video sequence recognition of transmission lines is easy to produce missed detection and false detection,kalman filter is adopted to track the transmission lines,and the connection between the front and back frames of transmission line images is established to improve the robustness of real-time recognition of transmission lines.(3)Set up the tower image data set.Aiming at the problem of insufficient data,several methods are adopted to augment data.This paper adapts YOLOv3 object detection model based on convolutional neural network.Anchor box is acquired through clustering algorithm to make it more conform to the shape and proportion of tower.This paper improves loss function,and calculates bounding box regression loss by GIoU.Experiments show that the mAP of the improved YOLOv3 algorithm in this paper reaches 90.8%,which has higher accuracy and faster detection speed.