Three-dimensional Information Acquisition And Defect Detection Of Wall Surface Based On Structured Light

With the development of society,the labor cost is getting higher,at the same time robot technology is making rapid progress,so using robot to replace the human in interior finishing task is a feasible way.This papers involves using a method to obtain threedimensional information of the wall,processing of point cloud and integrating the sensor into the robot system effectively.Measuring three-dimensional information precisely of the wall is the first requirement in an autonomous interior finishing task.Actually,the sensors used for this purpose should have large field of view,high precision and good stability.Existing technologies,such as stereo vision,Kinect,TOF(Time of flight)camera are not capable for automatic interior finishing robots due to the ill illumination condition,precision demand.Also,the sensor should be controllable and easy to use.Based on the analysis to the special requirements of the aforementioned sensor for interior finishing work,this research uses Gray code based structured light method to obtain the three-dimensional information of a wall.Since the projected fringe coding information is known in advance,this measurement method is applicable to flat and walls with less features.At the same time,for the possible changes in indoor lighting,this paper attempts to decompose the light to obtain a structured light measurement system with a certain robustness.The measurement system is consisted of a black and white camera and a projector.The projector generates a pre-produced gray code pattern,while allowing the camera to capture the target with projected code simultaneously,and the system then undergo decoding,calibration,triangulation and other steps to acquire the point cloud of the wall.For the point cloud data obtained,it needs to localize the position of the defect points on a wall.Taking into account the overall structure of the wall for the plane structure,plane fitting is a good way for this purpose.Then with this method the points outside the plane can be recognized as a defect point for the subsequent grinding.At the same time,in order to cope with the control of the robot arm,it is necessary to obtain the defect point position with respect to the robot coordinate system,which requires to calibrate the geometry relation between the structured light measuring system and the manipulator.In this paper,the external parameters of the camera are decomposed into the form of the matrix with the expected pose relationship,and the pose relationship of the final camera and the robot base system is obtained by nonlinear optimization.At the same time in order to obtain the overall structure of the wall from the points captured in different position,a simple point cloud registration is accomplished.In experiment,the structured light scanner can obtain one hundred thousand points stably.By analyzing the obtained point cloud data,it is concluded that the accuracy of the measurement system is about 0.42 mm at working distance of 500 mm.The defect localization accuracy in working distance of 500 mm can reach to 3mm.This accuracy can meet the general requirements of the rough wall polishing.These results may have some inspiration on the research of interior finishing robot.