Research Of Detection Method For Multi-measurement On 2-D Contour Based On Computer Vision

In modern detection technology, precision measurement of mechanical parts'dimension is a very common application of great practical value. With the unceasing improvement of the degree of automation, human vision detection or conventional contact measurement has dissatisfied the requirements of today's industrial fields. So, the development of non-contact measuring technology becomes more and more important. In this case, detection technology based on computer vision has appeared as a new method accompanied by the development of computer technology and photoelectric technology, which is a new technique based on modern optics in measurement area and also is a modern test technique including optical electronics, computer iconography, information processing and computer vision etc. The instrument and equipment on the basis of computer vision detection technology are able to realize intelligence, numeralization, miniaturization, network, and multifunction, and has the capabilities of inspection on real-time, real analysis and control. The technology has attracted wide attention in china.First, the dissertation detailed expounds research status of computer detection technology at home and abroad, explains current existing problems and development trend of computer vision inspection, and shows the direction of subsequent related research. It also discusses computer detection technology presents a high significance of theoretical research and practical application value as a domestic modern industrial detection method. Then, the dissertation explains the vision detection system components and discusses the three key factors influencing the system measurement accuracy: system calibration, image acquisition and image processing.Next, the paper profoundly studies related theoretic and algorithms of image processing such as image preprocessing, edge detection, sub-pixel location. In the collection, transmission and recovery process, image is inevitably interfered by various noises, and filter plays a role in decreasing noise. The paper detailed discusses mean filter and median filter by theoretical analysis and experiment, and takes the measure of combining the two filters to improve image quality. In addition, image edge detection technology is also discussed, which includes classic edge detection operator, canny edge detection operator, method of edge detection based on wavelets transform and artificial neural network edge detection. The advantages and disadvantages of the edge detection methods mentioned above are compared. Sub-pixel edge detection algorithm is one of the key technologies that improve the accuracy of visual inspection system by software. The paper explains the principle of sub-pixel localization and discusses typical sub-pixel localization methods mainly including fitting method, moment method and interpolation method, and on the basis a new sub-pixel localization method is formed, that is a universal algorithm for sub-pixel edge detection based on Sobel Operator which is not only universal for different kinds of edge, but also has good anti-noise ability and achieve sub-pixel localization greatly.At last, through consulting lots of foreign and domestic materials and seriously studying the core technology of modern detection and image processing algorithms, according to the size detection requirement of measured sample and its contour features, the paper takes the project of measuring IACV key size for Jilin Province JiRui Machine Electricity Equipments CD., LTD as support to design a detection system for multi-measurement on 2-D contour based on computer vision. The paper starts from improving measurement precision and speed by optimizing software algorithms, goes deep into studying the programming language of LabVIEW and image processing software IMAQ Vision, takes these as the platform finally to develop the entire system procedures and makes full use of advantages of LabVIEW such as software modularization, clear structure, short development cycle, strong flexibility and expansibility and easiness to maintain. System basically achieves detection mission about the key size parameters and analyses the error of measurement result. Accuracy of measurement can reach 0.003mm by comparing measured value with standard measuring instruments and the reasons for error are from systematic error and random error through analyzing.The end of paper summarizes research production and the work done, as well as suggestions on future work. The production of research can provide important reference for development of vision detection theory and the further study of sub-pixel positioning technology.