A Study Of Defect Detection Technology For Magnetic Parts By Machine Vision

In recent years, machine vision has obtained a rapid development and a wide application. Due to machine vision detection systems have several advantages such as non-contact, high accuracy, real-time, and on-line, the machine vision has a good application situation and develop prospects in industrial automation, medical testing, intelligent transportation, and many other fields. This dissertation is mainly concerned with the size and defect situation of a H type of magnetic parts used in inductance. Based at the analysis of research background, current situation of the development both at home and abroad, reading lots of references, carried on the thorough research of machine vision detection system, gives the specific solutions. The author’s major contributions are outlined as follows:1. The structure of machine vision detection system has been studied. In view of the two key indexes of speed and accuracy of the detection in actual project, the lighting, camera and lens technologies of the hardware module have been studied further. The hardware module is composed of MV-VD 200SM/SC type camera, AFT-ZML 0912 type lens and AFT-RL 12068 R type circular light. Using multi-sensor visual measuring technique with 3 cameras to get pictures, then deal them with the multi-threading technique.2. The digital image development tools have been studied. By comparing the effect of image processing and algorithm running speed of Matlab, choosing the open source computer vision library(Open CV), which has the features of open source and cross platform, as the development tool, and the compiled environment is Microsoft Visual Studio 2010, using C++ language as the development language to do the project.3. The digital image processing algorithm has been studied deeply. It contains image pretreatment process, image filtering and enhancement, edge detection operator and Hough transform for straight line detection and so on. For large defect parts, it is easy to find it by the perimeter and area, the difficulty of the thesis is for the inspection of tiny defect parts. For Circle detection, the requirements of the Hough transform are too idealistic to project, the polygon fitting and external fitting have been proposed to solve the problem. Combined with the actual process, the best algorithm is selected to achievethe detection purpose.4. According to the requirements, working process of the solution is established. At first, it is necessary to ensure the stability of calibration parameters of the camera calibration for the system. The design the working process. Finally, compile testing codes to realize the detection algorithm.5. Using the Statistics and analysis of the experimental results to verify the hardware and algorithms for the machine vision detection system designed solution can meet the detection target efficiently.