| With the expansion of urban construction scale and the acceleration of urbanrural integration,users have higher and higher requirements on the reliability of power grid.The operation and inspection of transmission lines is becoming more and more important for the normal operation of power grid.However,the number of professionals engaged in transmission line inspection has not changed,and the contradiction between the increasing scale of power grid and the structural shortage of inspection personnel has become increasingly prominent.The traditional manual operation and maintenance mode is difficult to meet the operation and maintenance needs of modern power grid,and a new intelligent equipment that can replace manual operation is urgently needed.With the development of robot and artificial intelligence technology,the research and application of transmission line inspection robot is an important way to change the work mode and improve the work quality.Aiming at the requirements of transmission line inspection robot line walking and target inspection,this paper designs a convolution neural network based transmission line obstacle detection and identification positioning algorithm and target level by level inspection algorithm.The main contents of this paper are summarized as follows:Firstly,a transmission line obstacle detection algorithm based on yn net is proposed,which can determine whether there are obstacles in front of the current line with the minimum operation loss.In order to meet the requirements of YNNet lightweight and stable and reliable,this paper designs the obstacle image sensitive region extraction mechanism,sensitive level division mechanism and reduce group convolution unit.Experiments show that YN-Net can achieve the best resource saving effect while maintaining the most stable detection effect.Secondly,a transmission line obstacle recognition and location algorithm based on Sep-Net is proposed.After YN-Net detects the obstacles in front,SepNet is activated to recognize and locate the obstacles in front.According to the characteristics of line obstacles and the performance limitation of embedded platform,the G-SN1 and G-SN2 feature units are designed,and the obstacle recognition and location network Sep-Net is constructed.To solve the problem of insufficient samples in the collected data set,feature extraction module and feature application module are designed for Sep-Net,and different data sets are used for combination training.Experimental results show that Sep-Net outperforms other similar algorithms in precision and recall,and keeps the best real-time performance.Then,a target patrol scheme based on Sep-Ne Xt and a step-by-step patrol algorithm are proposed,which is graded according to the characteristics of the patrol targets.The target patrol scheme is designed to patrol all levels of patrol targets in a progressive way.The step-by-step patrol algorithm Sep-Ne Xt involved in the patrol process is designed.In order to improve the patrol algorithm,the effective region cutting mechanism and detection anchor mechanism are designed.In order to verify the reliability of the patrol scheme and the performance of the patrol target detection and identification network,the comparative experiment,Ablation Experiment and field patrol experiment are designed.Finally,the hardware platform and software system design of the power line inspection robot are analyzed and introduced.The composition of the hardware platform and related physical diagrams are shown and explained.Design the composition of the software system and the communication mechanism between each module.The software and hardware coordination experiments are carried out to verify that the software and hardware coordination capability of the transmission line inspection robot in this paper can meet the practical application requirements. |
PDF Full Text Request