Research On Object Measurement And Identification Based On Machine Vision

The measurement and identification of object based on machine vision with its advantages of fast, precision and flexibility, have been studied comprehensively and used widely abroad and got more attention in the country. The thesis aimed to measure the dimension of workpiece and identify the power line from sequence images. In this paper, the machine vision technology is studied for the two types of object.Identifying and measuring the workpiece images. Firstly, according to the feature of the workpiece with smooth and highly reflective surface, this paper design LED back lighting source system with uniform scattering and it can obtain the high contrast of target and background, clear edge contour image. Secondly, the image preprocess is analyzed on object image; noise is mainly reduced by Gaussian filter, and the OTSU method and Canny edge extraction approach are designed to achieve the image of binaryzation and extract the edge. Thirdly, the binary centroid determination of image method and the improved Hough transform circle detection algorithm are designed to complete the target feature point extraction. Finally, relative calibrating method is used to achieve a camera calibration, then the measurement of the size of workpiece is completed and the cause of the error is analyzed.Tracking recognizing power line on aerial sequence images. First of all, using the enhancement algorithm of bilateral filtering linear goals to enhance linear objects. Then, parallel line recognition based on Hough transform is carried on the enhanced results. Eventually, according to the correlation of sequence images, after estimating the current state and limiting the scope of the power line, tracking and recognition are done on power line in sequence images. On the basis of the traditional image enhancement and recognition, direction constraint and angle constraint are respectively increased, which improved the recognition efficiency significantly.In this paper, the measurement and identification system based on machine vision can accurately measure the size of workpiece information and recognize the powerline from complex background image. Experiment shows that the improved Hough transform circle detection and line detection algorithm can not only meet the requirement for measuring accuracy, but also largely improve the detection efficiency, so it has a certain theoretical value and application value of engineering practical.