Research On The Search And Identification Of Power Facility Targets That Can Be Used For Automatic Cruise

With the rapid development of computers,artificial intelligence is becoming more and more common.Intelligent housing,intelligent hospitals,intelligent industries,intelligent agriculture and other phenomena have emerged,and the power industry is no exception,the scale of overhead transmission lines is gradually expanding.PoIes,towers and power lines are exposed to the wild for a long time,which are vulnerable to natural erosion and man-made damage.A series of problems,such as line aging,insulator damage and anti-vibration hammer loosening,will bring incalculable losses to the entire power system if not repaired.Traditional manual circuit inspection is not only a heavy workload,low efficiency and difficult working environment,usually maintenance personnel to cross mountains,rivers and other complex terrain,more importantly,manual power inspection will affect the daily operation and maintenance of the line.However,in the event of natural disasters such as landslides,earthquakes,rain and snow,the uav can normally conduct on-line inspection with high efficiency.Up to now,there is no real fully automatic uav power inspection system in the market.Aiming at this problem,this paper conducts an in-depth study on the detection and tracking of three typical power targets,namely power transmission tower,insulator and power transmission line,to prepare for the next step of uav automatic inspection of power facilities.Aiming at the problem of transmission tower detection and tracking,this paper studies the detection and tracking algorithm of transmission tower.Mainly for reinforced concrete tower tower,steel tower and the three material transmission tower for detection and tracking,according to the cross of tower circulation structure,steel tubular tower and linear characteristics of reinforced concrete tower,respectively to extract significantly pixels in the image of the ORB,influence on the local deformation and light has a good stability of HOG feature and image gray-scale changes such as illumination change has stronger robustness characteristics of LBP training multiple classifier model,and measured by the test set extraction pylon HLORB(ORB,HOG and fusion of LBP features)characteristics,The XGBoost classifier in training integrated learning has the highest accuracy,reaching 96.58%.It USES a selective search algorithm to search the transmission tower,and it takes 42ms to detect an image.Based on the detected transmission tower,KCF algorithm was used to track the transmission tower,reaching 33 frames/s.Aiming at the problem of insulator identification and positioning,a deep learning algorithm is proposed to detect insulators,that is,the improved algorithm based on YOLOv3 is adopted to detect insulators,and MobileNet v1 is selected as the feature extraction network,with the recognition accuracy reaching 98%and the detection rate of 29 frames/s,realizing the real-time detection of insulators.To solve the problem of power line detection and tracking,a power line extraction algorithm based on LSD is proposed.Aerial photo image blur due to jitter of unmanned aerial vehicle(uav)in flight,through the comparison of various filtering algorithm,the image edge of bilateral filtering algorithm to filter out the noise of the image,Sobel operator to extract the image edge,compared with hough transform,LSD,straight line segment fracture were detected,fast speed and stability,less use of LSD Algorithm to extract image line,will have to extract the main direction of straight lines are extracted,then according to the length of the straight line segment selection,finally using the least squares fitting lines and able to identify 92%of the power line,the use of nondestructive kalman filter for power line tracking,rate reached 32 frame/s,realize the tracking of video sequences.At the same time,the uv channel image and visible channel image of the sunblind uv dual-channel detection system are integrated,so as to make preparations for detecting corona discharge intensity of power facilities and locating defect position of visible channel in the next step of actual power inspection.