Research On Target Detection And Visual Servo System Of Autonomous Inspection Robot In Substation

With the rapid development of power system intelligence,replacing humans with robots to complete some special daily inspection and management tasks has gradually become a new trend in the development of my country’s power system.This paper designs and develops a substation autonomous inspection robot control system based on target detection and visual servoing algorithms for the needs of indoor and outdoor inspections in substations.The main research work is as follows:The overall requirements of the automatic inspection robot control system for substations are analyzed,the overall design plan of the wheeled inspection robot based on the hierarchical design idea is proposed,the selection of related core equipment is completed,and the software architecture of the robot is described.The internal and external parameter model and distortion model of the camera are established by using the pinhole camera model.Analyzed the vision system model of the robotic arm with the eye in the hand,using the Tsai’s two-step method to solve the hand-eye relationship matrix,and using the Zhang’s calibration method and the Tsai’s two-step method to simultaneously calibrate the internal and external parameters of the binocular camera and the hand-eye relationship matrix.Aiming at the problem of uneven light distribution in substation indoor inspection scenes and difficulty in identifying the state of the knob switch on the electrical control cabinet,an improved YOLOv3-Tiny-based knob switch position detection and state recognition algorithm is proposed.Relevant parameters were adjusted on the original YOLOv3-Tiny network,deduplication and sorting algorithms were added,and the self-made data set was used to train the network.The final design experiment verifies the robustness and real-time performance of the algorithm in target detection and target state detection.According to the description of the visual servoing task of the robotic arm in the indoor inspection of the substation,the kinematics model of the robotic arm is established,and the IBVS system is selected to realize the visual servoing of the robotic arm.In view of the narrow aisles in the indoor scene of the substation,the trajectory planning of the robotic arm on the mobile robot may have workspace constraints and field of view constraints,a robotic arm visual servoing algorithm that weakens the field of view constraints is proposed,which is based on an improved TLD target The tracking algorithm is designed to first extract the image features of the target object in the current frame(objects in the camera’s field of view)or predict image features(objects outside the camera’s field of view).Then further estimate the position of the manipulated object in the current image.Finally,a visual sliding mode control law is designed to control the motion of the manipulator according to the image feature error,so as to complete the visual servoing task of the manipulator with high robustness when the target deviates from the field of view or the target is blocked.The target detection and visual servoing experiment environment of this article is built according to the indoor scene of the substation on campus,and the test and verification of the target detection function of the substation autonomous inspection robot and the visual servo function of the manipulator are completed,and the feasibility of the algorithm is verified.