Font Size: a A A

Research And Application Of Image Measurement Technology In Object Size Detection

Posted on:2016-05-19Degree:MasterType:Thesis
Country:ChinaCandidate:H XuFull Text:PDF
GTID:2308330464971906Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
The image measurement technology is based on digital image processing, combined with optics, circuit, signal processing and computer vision processing etc., which is a highly performance non-contact measuring method developed in recent years. Rather than the field of image measurement, image measurement also has a wide range of applications such as image pattern recognition, image matching, image positioning. Compared with traditional measurement method, image measurement has a high measurement precision, wide range of applications, great controllability, and many other advantages. In this paper, we study the algorithm of image measurement based on the principle of digital image processing and validate it through multiple experiment results.The image measurement process is mainly contained with image preprocessing, image binarization, edge detection, linear/circular detection and parameter calculation. Image preprocessing includes image enhancement and image denoising. Image enhancement is based on gray-scale transformation enhancement and histogram enhancement, by improving image contrast to improve the identification degree of the target. Image denoising is achieved by filtering, which contains spatial filtering and frequency domain filtering. In this paper, we use median filtering in spatial filtering to eliminate the noise in linear detection which also has the effect of edge smoothing. We adopted the homomorphic filtering of frequency domain to reduce the influence of illumination for uneven illumination of the image in circle detection part. In the part of image binarization, we used the between-cluster variance method (OSTU) in global threshold method to divide the target and the background of the image, which can adapt to image measurement to achieve the different requirement of division. We use the Canny operator in edge detection, which has a higher accuracy and anti-noise performance than Sobel operator, Roberts operator, the LOG operator and other first-order or second-order operator. We use Hough transform algorithm in line detection, by using original image space and parameter space transform to extract straight lines in the image. Round detection contains full round and incomplete circle detection. Both of them, we use the least squares fitting method, which has a higher accuracy. By changing the number of starting and connecting point, we acquire the average result of multiple experiments in order to reduce the error of measurement.In our experiment, we use line detection, circle detection and parameter detection in complex shape object through several pairs of real objects and analyze the reasons of the error in our measurement results. The experiment result shows that, the measurement process and algorithm in this paper are feasible. Under the non-ideal circumstance of non-professional camera system and environment, the accuracy can still reach the level of millimeter, which can meet the requirement of most measurement system.
Keywords/Search Tags:Image measurement, Image binarization, Edge detection, Line detection, Hough transform, Least square fitting
PDF Full Text Request
Related items