Study On The Recognition Algorithm Of AGV Dynamic Path Based On Machine Vision

With the development of modern industrial automation, versatile automatic tridimensional storehouse has dominated various industrial fields. As being the link and the terminal that adjusts the flexible logistic system, Automatic Guide Vehicle (AGV) represents an iconic symbol of modern continuous logistic transportation. In order to enhance the accuracy, timeliness, and adaptation of visual guiding AGVs, to achieve more agile guiding and efficient path plan as the final result, the thesis focus on the continuity of adjacent frames, analyzed and studied the navigation lines, marks and obstacles in the project.By studying the current development of AGVs in other countries, analyzing and comparing the operational principle of various guiding/navigation system among AGVs, the thesis stated superiorities and flaws of the visual guiding AGV system, hence gave a clear view of what needed to be solved from the thesis.Via the establishment of mathematical modeling on video camera system, which included the setting of coordinated system and analysis of lane module; during the image pretreatment, and binary processing of gray scale image based sequential images; and examined the margin of navigation mark lines, acquired the dynamic data of the path environment during AGV's operation. In order to increase the accuracy while AGV was tracking the route, studied the testing method of central line of guide mark, and found that conventional Hough transform algorithm took much longer time while extracting the guide central line. Consequently, studied the Hough transform algorithm that based on pretreatment, and extracted the central guiding line. As the result showed, that improved Hough transform algorithm enhanced the timeliness of entire system, proved validity of improved algorithm.On account of lacking in enough data from the simplex navigation lines that in practical operations of a visual guiding AGV, and for the purpose of improvement, enrich the path information and enhance the intelligence of the system, there were two sorts of simplified guiding marks that designed for the system identifying, they were:control marks and digital marks. These two sorts of marks were both treated by image projection that projected to the interested area, then identified the marks according to their own structural eigenvector. The result of experiments exhibited the accuracy and practicability of the algorithm.To fulfill the safety requirement of AGVs, and to make space orientation of the obstacles on the path. For detecting and tracking the obstacles in sequential images, it started from the single image frame by using ground plane technique to determine the obstacles; then, via feature fusion, using mean shift algorithm to track obstacles; and applying examining and tracking interaction tactic to optimize the entire process; at last, according the cross ratio invariability principle of projective geometry, determined the space orientation parameters of obstacles from known parameters, as well as corrected the errors. All the experiment data showed that interaction between examination and tracking mechanism was able to promote the accuracy of obstacle detection, and space orientation algorithm was able to achieve a higher accuracy and timeliness.