Research And Implementation Of Several Key Technologies In Automatic Guided System Based On Computer Vision

The automatic guided system based on computer vision is primary one of computer vision engineering application. In these years, it is used widely in application, because of its high guided accuracy, cheap price, and the ability that it can adjust the route automatically with the accident.This dissertation studies three key technologies in automatic guided system based on computer vision, which is the extraction, description and tracking of the guided objects. The research findings are used in two projects that UAV automatic reclaim system and automatic guided vehicle system. And these research findings are well and valuable for both military and engineering application.Firstly, the dissertation summarizes the common characteristics of the automatic guided system based on computer vision, and provides the design means. Otherwise, it also provides the design plans and mathematics modelings about the two projects above.Secondly, the extraction of guided objects is researched. In edge detection, a new directional morphological gradient operator (DMGO) is proposed which uses gradient projections to both provide orientation with also to reduce sensitivity to added noise; In the segmentation, it includes a new improved optimal threshold (MSOT), which makes edge more intact and clearer; In image repair, a new simple arithmetic about jointing broken lines was showed. It doesn't only make the targets'contour smooth, but also can find the single pixel boundary with tracking.Thirdly, the research about the description guided objects. To address the shortcoming of current object regional descriptors, a new set of object regional descriptors based on moments is provided in this paper. The descriptors are named EED (Enclosing Ellipse Descriptors). It uses the properties of moments and the multi-scales of binary tree, cuts the object region into several parts, and draws every part of the object regional enclosing ellipse. Then, these obliquities of enclosing ellipses are used to describe the object region. This set of descriptors is not only invariant to translation, rotation, and scale change, but also is invariant to detail difference. The detail difference doesn't affect the recognition rate, if only the rough shape of objective region is not change. The five properties are proved by the means of theory or test. At last, the descriptors are used in the actual projects on automatically guided system.Fourthly, the technique of the tracking of guided objects is studied. The dissertation points out that the automatic guided system has two performance parameters, speed and accuracy. A mathematics modelings of windows tracking is setted up, a integrated resolvent plan is provided, which can be used in the automatic guided system based on computer vision.Finally, this dissertation design and realizes the UAV automatic reclaim system and automatic guided vehicle system. All research findings are used in the two projects. In automatic guided vehicle system, a novel algorithm for automatic guided vehicle is designed. The algorithm uses the CCD in front of the vehicle to identify the routing indicator and to calculate the yaw angle from track. It applies the odometer to get vehicle displacement. The DR algorithm can provide the vehicle direction and position from the yaw angle and displacement. Machine vision guidance can identify and track the routing indicator.The research work in this dissertation will push on the application of the automatic guided system based on computer vision more deeply in engineering.