Embedded Dimension Measurement System Based On Asymmetric Binocular Machine Vision Technology

The traditional method of dimension measurement is based on monocular machine vision technology, which gets the complete information of the objects to be measured by using a single image, and then processes the image to obtain the dimensions. But when the objects to be measured are the local and global dimensions of the large complex parts, or their geometric dimensions are changed greatly, it's hard to accurately measure some of them because they are too small. Now, most of the traditional measurement devices are still based on PC, which are expensive and bulky. It's hard for them to meet the demands for industrial measurement and control. Therefore, the study on embedded dimension measurement method and device has important practical significance.The purpose of this thesis is to design and implement an embedded dimension measurement system based on asymmetric binocular machine vision technology, which combines the features of the binocular and monocular machine vision technologies. The whole measurement Process is achieved in the software and hardware platform based on ARM, and the operator can remote manage and control the device by the browser of the client. The major work is as follows:1. The measurement method and the dimension measurement procedure of the dimension measurement system based on asymmetric binocular machine vision technology are put forward. An embedded implementation based on ARM is designed, and the principles of choosing the hardware are proposed.2. Feature extraction algorithms are studied, and in order to improve the speed and accuracy of the corner and edge detection methods, a corner detection algorithm and an edge detection algorithm are proposed. And on this basis, system calibration is achieved by using camera calibration method based on plane grid point.3. With the actual of the cameras of this system with different focal length lens, a stereo matching algorithm is designed according to several constraints. A method to determine the plane is proposed, and after determining the plane of the object, the dimension of the object can be measured by just one camera.4. An implementation of the Embedded Dimension Measurement System based on ARM is developed. The hardware structure and software modules in WinCE operating system are proposed. The functions of remote monitoring and control are implemented by using the technologies of COM and ASP. The results of round-parameter measurement experiment show that this embedded system is able to achieve the expectations.