Research On Feature Detecting Technique Of Linear Movement Trajectory For Robot's Performance Calibration

Industrial robots are gradually evolved into main link on production line in factories, but the full realization of this goal is inseparable from the performance test trajectory. With the increasing accuracy of operation, this technology has become bottleneck in using industrial robots smoothly and efficiently. Currently, such measuring equipment in foreign countries has reached the sereotypes of product, and domestic are still depend on the import. In this situation, it is selected as main objective to work for.The category of task is structured-light vision sensor non-contact measurement for linear trajectory of robots. Using structured-light geometry imaging principle and vision image processing technology, through online image collecting and off-line processing, the measurement for linear trajectory of robots is realized. This paper concerns solutions of key difficult problems during the research. First, when analyzing the advantages and the disadvantages and potential space to develop of each processing method thoroughly, the reasonable and feasible frame and technical implementation route are established based on the overall analysis around the field of present technological development. Second, beginning with image processing, all kinds of difficult technology problems of system software identifications and extract reference features, which follow with measurement system image collection, are analyzed and solved one by one. At last, based on the massive experimental results, testifying system effect and describing realization technique are given.This paper tells of the systematic countermeasures of many difficult problems, including image processing principle of line-structured light strip, image enhancement of structured light strip, image segmentation of structured light strip, thinning, fitting lines, identifying the turning points of structured light strip image and removing burrs, etc.. At the same time, it presents an automatic multi-threshold image segmentation and image regional dynamic division technology, which is proposed by the author. The use of automatic multi-threshold image segmentation, can not only overcome with satisfaction that the uneven lighting, or uneven distribution of gray on the impact of target recognition, but also a good solution to the region from neighboring threshold "step" caused by the wrong division, which makes the final system has an extremely excellent robustness. Image regional dynamic division is based on regional dynamics of grayscale distribution of images, ranging from the number of its divided into several blocks, each block choose different thresholds. In this way, it is not only a good solution to the changes in light conditions automatically division of the image, but also the elimination of the same light on the different locations because of the uneven distribution of gray caused by the impact of image segmentation. In the end of the paper, the experimental results of the system are given, and pointed out the lack of surviving and future research directions.