Research On Application Of Visual Localization Technology For Food Testing And Weld Grinding/Polishing

With the development of production and the improvement of the industrialautomation, the robot vision becomes more and more important in the automaticproduction. The technology of visual localization is one of the most active branch inthe research field of robot visual, and it is also an important symbol in thedevelopment of the intelligent robots.Based on the previous studies and large amounts of data, this paper utilized thetechnology of visual localization to recognize and locate target by the robot. Firstly,the online food testing platform structures were introduced, a platform carrying foodon a conveyor belt for visual orientation, visual localization technology was utilizedto locate the food on the conveyor belt. The paper introduced the overall layout andfunction of the platform and explained the computing process of the motor servomotor selection and drive type comparison between different models. Then thealgorithm for visual localization was introduced.Secondly, the paper studied the visual localization technology and applied it tothe weld grinding and polishing robot, the purpose was to provide positioninformation and height of the seam for the weld grinding and polishing robot, thushelping the weld grinding and polishing robots performed the automatic grinding andpolishing task. The process of visual localization of weld seam mainly consisted ofthe following parts: CCD calibration part, reference origin parts, image pre-processingpart, feature extraction part and three-dimensional measurement part. This paperfocused on the pre-processing and feature extraction section of weld seam image. Inimage pre-processing part, the paper used the methods of gray scale maximizing andmedian filtering to process the image briefly, and changed the gray scale maximizingmethod based on subsequent processing steps, in other words only used it near theedge of the weld seam. In the feature extraction part, the paper introduced twodifferent methods of extracting the center line, extract the different characteristics of center line by comparing the average value methods of line coordinate and normaldirection, meanwhile introduced the method of feature point extraction based on thedistance.Thirdly, based on the software development of weld grinding and polishing robot,the image processing underlying module was programmed in the HALCON softwareenvironment with the method of hierarchical modular, which mainly included imageenhancement module, calibration of CCD module, three dimensional measurementmodule and the feature extraction module. With the help of underlying module, thefunction integration of middle module such as calibration and image enhancementcame true. The weld grinding and polishing robot software interface was developedunder the VC++environment. Combined with the interface visualization of VC++program, the rich function library of HALCON, the visualization of visual system wasoperated, and the realized the integration of the top module, thus making theautomation of welding seam image processing possible.Finally, based on these studies, the paper carried out multiple experiments toverify the performance of the algorithms, including visual localization of biscuits onthe conveyor belt, the area, the center coordinates and pose parameters of the biscuitswere measured. For the weld grinding and polishing robots, first was to verify thevalidity and stability of the visual system, then conducted experiments on lightinterference and vibration interference, and analyzed the experimental data, then madesome improvement measures. Experiments showed the weld visual localizationsystem had certain accuracy and stable function, was able to provide reliableinformation to the weld seam grinding and polishing robot.