With the development of hydropower industry,the demand for quality inspection and maintenance of large-diameter penstock is increasing.Climbing robot can replace human to inspect and maintain the surface of penstock,which greatly improves efficiency and safety.Aiming at the inspection and maintenance of penstock in Three Gorges,the hardware platform of climbing robot is built.The localization algorithms and control system of climbing robot are designed by utilizing the knowledge of motion control,image processing and software engineering.In order to adapt to different circumstances,trajectory estimation algorithm based on wheel odometer and visual localization algorithm based on monocular camera are designed respectively to realize the localization of climbing robot.In the visual localization algorithm,an image clipping algorithm is designed to select the region with best quality in image.The number of feature points in image is increased by equalization.The realtime position and pose of camera are obtained by key frame tracking and local map tracking,which avoids the position and pose of camera fall into local optimum and improves localization accuracy and robustness.The control system of climbing robot includes the user software in upper computer and the real-time motion control system in lower computer.The user software is developed by C# WPF MVVM model.The lower computer realizes the motion control of climbing robot by Beckhoff controller.The control system and visual localization algorithm of climbing robot were verified by experiments.The control system realized the realtime motion control of climbing robot.The visual localization algorithm was with high accuracy and stability.For most penstocks,the visual localization algorithm could extract feature points accurately and locate robot in real time. |