Recognition And Fault Detection Of The Key Components In Transmission Lines Based On Aerial Images

The electric power line inspection of unmanned aerial vehicle(UAV)is adopted by more and more power supply companies for its convenience,clear acquisition images,manpower and material resource economizing.The UAV controlled by the inspector flies along transmission line and photographs the electrical equipments along the way by the camera it carries.And then the recognition and fault detection of the related equipments are realized by image processing and machine learning techniques.This application further promotes the construction of smart grid.In this thesis,the recognition and fault detection techniques of two key components(insulators and stock bridge damper)in aerial images are studied,which provides the theoretical basis for further automation of UAV patrol equipments.Firstly,this thesis introduces the current research status of the identification and fault detection of key transmission line equipments,and analyzes the current situation,advantages and disadvantages of deep learning algorithm.The depth learning algorithm is based on large database,but the real aerial images can not meet the requirements of quantity and quality in deep learning.In order to solve this problem,this thesis proposes a method to simulate artificial images in depth learning.The 3ds Max drawing software is applied to the establishment and expansion of the image sample library.The 3D model of insulator and stock bridge damper is drawn,and the artificial samples are obtained.Besides,in view of the inaccuracy and low recognition rate in object positioning of existing deep learning algorithms,this thesis puts forward a regional proposal method based on the color name,which ensures the classification accuracy and improves the recognition accuracy.The final recognition accuracy of the classification model in this thesis is 91.7%,which is increased by at least 5 percentage points compared to the other methods.Subsequently,a fault detection method based on adaptive morphological feature model for insulator bunch-drop failure is proposed.The method is advantage with superior rapidity and accuracy.The final fault detection rate reaches 87.3%.