Image Recognition Of The Flexible Manipulator Pose

At present, flexible manipulator has been extensively used in various fields, such as mechanical electron, aviation, micromanipulation etc. As well as it has been a hot topic in research[2] . A lot of researchers make much research in flexible structure , flexible material , control , dynamics model etc, but obtaining the flexible manipulator pose is seldom mentioned . It is exactly a very challenging subject in robot to combine vision technology and manipulator control technology[11], Image recognition of the flexible manipulator pose presented is exactly a exploration of this field. Researching discernment of flexible manipulator pose in real time based on binocular stereo vision, we discussed some key problem including the establishment of the binocular stereo vision system, the calibration method of paramenters for the binocular stereo vision sensor , the stereo matching method, the ecognition flexible manipulator and pose mensurement. The major research content is as follows : In chapter 1, the mission and contents of research project are presented. And difficulty to research project is analysised. The mission of the subject is to set up the real time image acquisition and processing system based on binocular stereo vision, recognize flexible manipulator in real time and resolve the pose of flexible manipulator. The measurement velocity of the system is 60 ms per frame. The relative error of resolving the pose of flexible manipulator is demanded below 5 percents. Mismatching rate is below 10 percents. The algorithm of recognizing and stereo matching, algorithm implementation are difficulties in research. In chapter 2, the binocular stereo vision is introduced at first. Then the concepts of computer vision and stereo vision are gived. Especially, the principles, characteristics, classification, development, application of stereo vision are introduced in details. On the basis of the actual situation of research project , the binocular stereo vision system is set up, which is natural light source, eye-to-hand. Two fixed cameras, which optical axis are is parallel, installed horizontally. The chapter 3 deals with the key technology of binocular stereo vision—setero matching. The content, methods of setero matching are introduced. The advantages and disadvantages of various methods are discussed in details. Then stereo match edge line based on line support region is carried out. The algorithm not only exploit location, direction and the length of edge line, but also exploit the color information contained in the image, it make the matching more objective and accurate. The algorithm has advantages of feature matching and area matching. For better degree of matching, stereo match edge line based on fuzzy is presented. Through analyzing complex background and simple background two conditions, the paper gives the degree of matching The chapter 4 deals with the vision system's calibration, the image processing, recognizing flexible manipulator and resolving the pose of flexible manipulator. The calibration is carried out off-line with the traditional linear model, which is simple, and then the relation between the image coordination system and the world coordination system is obtained. Edge points are made easily detected and distinguished with modified laplace operator. Based on natural acknowlege, a recognition of flexible manipulator acknowledge-based is proposed. Then , according space geometrical constraint, calibration and setero matching result, the pose of manipulator can be easy geted. The errors of system are listed in the last of chapter. The focus of the chapter 5 is on completing hardware and software design of the real time image acquisition and processing system based on binocular stereo vision. The system consists of two SK-802 CCD cameras , SAA7111, 1K50 and LPC2106. PAL video signal characteristics are analyzed; Design of the video decode circuit of SAA7111 chips with online parameter adjustment and the hardware design and debug of LPC2106 are interpreted; ARM series processors are introduced in details; in software design, the features of ARM, flow chart, program languages and data struct are shown. The setero matching algorithm is also optimized for LPC2106, making the module finish processing a image in about 60 ms per fram. In chapter 6, some conclusions of the paper are given, as well as the paper need to improve further. The main purpose of the research work is to measure pose of flexible manipulator in real time based on binocular setero vision. This purpose approach should be useful for many applications where convenient,low-cost,static and/or...