基于机器视觉的黑片缺陷检测图像特征提取技术研究-计算机软件与理论专业论文.docxVIP

沈阳理工大学硕士学位论文 Abstract Machine vision simulates human’s vision by a computer,in which a very important part is image processing.Feature detection is an important research aspect in image processing while corner is a very important feature of the image. The exactness of corner detection affects the understanding and analysis of the image and plays an important role on the determination of objects in the scene and three-dimensional view matching. In this paper,studying in-depth based on gray level and edge Corner Detection algorithm, an improved Harris multi-scale corner detection and corner detection based on integrating image feature were carried out based on multi-scale ideas. Harris corner detection is a classic corner detection algorithm. It is broadly applied in reality. But it doesn’t have the scale change characteristic. Even we change the parameters in image corner detection we can not get satisfying effect. in this paper, we introduce multi-scale space and the fuzzy parameter into the algorithm,detecting corners combining multiple-scale and Harris algorithm, filtering corners through a global threshold in finally.Combining the algorithm of this paper and the classic algorithm in one procedure helps comparative study. It not only includes the feature information at several scales, but also vercomes the drawback that the single-scale Harris detector usually leads to either missing significant corners or detecting false corners. Corner detection based on edge is of high precision but it is sensitive to noise and computing complicated while method based on pray-level is easy to implement but effects are not so well. So in this paper,a algorithm is proposed in which edge feature and gray-level feature are contributed to corner detection.First we use Canny to work out all the edge points at a low scale, and then work out the value of the curvature of each edge point and derive the initial set of corner points.Finally we determine the final set of corners with Harri