Machine Vision Defect Detection Based On Wavelet Transform And Genetic Algorithm

In recent years, the product detection technologies based on machine vision have widely been applied for its high efficiency, high reliability and low cost. This thesis has systematically studied the related techniques of machine vision products, put forward the specific machine vision defect detection methods on glass image, and design and implement a glass defect inspection system for automatic classification and recognition of glass defects through researching the related technologies on feature extraction and pattern recognition.The work of this thesis mainly includes:1.The concepts of machine vision and its application situation are introduced, the machine vision technology is applied to detection of image defects on the glass image.Due to the characteristics of the low image contrast and edge blur on the glass image, several selected for glass image pre-processing methods are used included gray-oriented processing, spatial linear transformation, binarization method and the wavelet noise removal.2.This thesis uses wavelet network, genetic algorithm, and K-L combination in the glass defect image feature extraction. Twice compression and one reduced dimension methods on the defect image is used to achieve a fast convergence while avoiding to fall into the local accumulation of small.This method makes the image feature extraction is more rapid and achieved good experimental results.3.This thesis presents a new classification model of WMDC--weighted minimum distance classifier. It expands the minimum standard Euclidean distance classifier scope of application and increases the minimum distance classifier the classification accuracy through sample attribute distance-weighted definition of the scope and increasing the property value constraints.The images inspection system is made up with the camera, image acquisition card and the composition of micro-computer. This involves only the software part, by using Visual C+ + defect detection of glass to complete the software development. Through theoretical study and experiment, detection systems can be used to detect glass bubbles, inclusions, defects such as nodules.Presented results in this article have proved that the proposed glass defect image detection methods are effective and feasible for the glass defect image machine vision inspection and laid a good foundation for further research and development.