Research On Visual System Of Teleoperation Manipulator

In order to monitor the status of interior scene or internal components under the hostile environment(high temperature,darkness,vacuum,nuclear radiation,etc.)where is dangerous or inaccessible for human beings,machine vision is always used to replace eyes for damage information.A visual system that installed on the end-effector of manipulator is a good choice.In this paper,our vision system is mounted on the end-effector of flexible in-vessel inspection robot(FIVIR)for full volume and close visual inspection of first wall in the inner chamber of tokamak,which providing the basis for the follow-up maintenance work.Centered on the function implementation of visual system,this paper mainly solve two problems.Firstly,it is non intuitive and tedious if check the damage by looking over the images sequence or videos,to make it more intuitive and improve the efficiency of visual inspection,this paper propose a method of first wall scene emersion based augment virtuality.It is designed for convenient,intuitive and realistic-looking visual inspection instead of viewing the status of first wall only by real-time monitoring or off-line images sequences.Based on the idea of augment virtuality,this method includes two main parts: first wall tile recognition algorithm and first wall tile image registration algorithm.The first wall tile recognition algorithm using robust image process algorithm is used to recognize and extract the tile unit image from the original images captured by camera and the first wall tile image registration algorithm is used to overlay extracted tile image onto virtual first wall scene model in 3D virtual reality simulation system by texture mapping way utilizing pose information of robot arm and estimated camera pose.Through this method,users can directly observe the real tile images in the 3D virtual environment of tokamak cavity,which have the advantages of strong three-dimensional stereo feeling,convenient and quick positioning of damage tiles.Secondly,since the cameras and light source are all mounted on the end-effector of FIVIR,camera tends to be exposure excessive or insufficient when observation-distance vary in the motion of FIVIR under dark environment.To solve this problem,this paper put forward a automatic lighting method based images brightness quality analysis to adjust the light value level automatically for good image brightness quality when observation-distance changes.This method works when the observation distances are available.We firstly establish the image brightness quality evaluation model for image brightness feature extraction by ways of feature learning,then establish the regression model among image brightness quality,observation-distance and light value.Take the designed image brightness quality feature and observation distances as the input of trained regression model,then we can get the predicted light values for good image bright-ness quality corresponding with the input observation-distances.Thus,we can set the light value according with the observation-distance real-time in the motion of FIVIR.Finally,we carried out the practical experiments which validated the feasibility and efficiency of the first wall scene emersion method based augment virtuality and automatic lighting method based images brightness quality analysis.