A Study Of Image Processing Methods For Robot's Linear Trajectory Measurement

Robots' linear trajectory characteristics are important performance indexes of industrial robots, in which repeatability and stability of robots' movement trajectory synthetically reflect robots' mechanical and electrical performance and the ability of controlling movement trajectory. They are the most important check indexes after robots have been made. But there have been being no robot trajectory measurement devices in our country, and they have been relying on importing for long time. Therefore, it is very important and urgent to develop robots' linear trajectory measurement devices.This thesis reviews the trajectory measurement systems of robots at first, then stresses on describing the non-touch robots' linear trajectory measurement system that is made in Shenyang Institute of Automation Chinese Academy of Sciences. This system is on the base of structured light geometric imaging principle and vision image processing technique. The vision sensor is fixed on the end-effector of the robot. When the robot makes a linear movement along the "V" measurement track, the linear light from the laser electropult on the sensor is projected on the track to shape laser strip with three turning dots and imaged in CCD camera. The relation between the sensor coordinate and the measurement track coordinate can be computed by image processing and pose measuring, which indirectly describes the characteristic of the robot's linear trajectory.In order to improve the precision of image processing, one of key techniques of this system, this thesis studies well robust technique of image processing and the feature dots recognition in the different measuring surroundings, which using the real pictures from the measurement system according to the technique standards. Firstly, image processing software based on BMP in Visual C++ is made, and it has the functions of reading, showing, processing and storing pictures. Secondly, image enhancement technique based on linear filtering is adopted. A new image segmentation method by means of automatic multilevel threshold is given, which realizes partial multilevel threshold segmentation based on image region partitionof gray-level position projection, removes the influence of uneven illuminance or uneven gray-level distribution on goal recognition, and resolves the problem of error segmentation caused by threshold step between adjacent regions by threshold transition. The method has well robustness. Thirdly, in the detecting of the feature dots in the images, the technique gaining laser strip skeleton images is developed, and it includes image thinning, boundary following and dealing with thinning image lastly. During the course of laser strip skeleton images fitting, the problem of adaptive segment segmentation is resolved, and well robust line segmentation method is present. Lastly, the 3D coordinates of the feature dots are given. Many experiments show that the methods can improve the precision and robustness of image processing.