Research On Digital Image Auto Focus Technology That Based On The Vision Measures System

With the development of image processing compared with computer technology and detection technology, formed the visual detection system. Use the visual detection system we can measure the small size goods with a high precision. But if we use other equipments we can not finish the precision measurement. Auto focus is the key technology in visual detection system, and auto focus is the foundation of image processing, pattern recognition, precise positioning and so on follow-up operations. Only quick and accurate acquire the clear image, we can complete the subsequent work effectively.In this thesis we take the visual tool detector as the research platform, around the technology of auto focus, research on the technology of clarity-evaluation function, focus search algorithm and region selection. Then we get the better algorithms that are suitable for the tool images, which include the following parts:1) In this paper, the source of the subject and background are described in detail and the development of auto-focus on the current status is introduced. The paper also gives a detailed introduction of basic theory principles which are related to the subject.2) Analysis the relevant theory of auto focus algorithm, including three modules:the clarity-evaluation function, focus search strategy algorithm and the effection of the region selection to the clarity-evaluation function. The clarity-evaluation function is evaluated for the tool image, judge the image is focus or defocus. It is a reference to the image quality. In this paper, we analysis and compare the common clarity-evaluation functions. The results show that the Laplace (8 neighborhood) has a higher precision in the tool image processing. The purpose of focus search algorithm is to achieve the searching and positioning of focus. Then quickly find the accurate focus position. In this thesis we analyze the common focus search algorithms, describe the flow chart of each algorithms. The results show that the "three point method" which improves from the hill climbing algorithm can avoid the local peak caused by error and this method has a high precision. The purpose of the region selection is to select the interested regions. Only analysis the interested regions, we can reduce the amount of computation. This thesis analyzes the advantages and disadvantages of each region selection algorithm, and then compared with the whole image. The results show that the region selection of more point window has a good effect, high computational efficiency. If we analysis the whole image, the data is large and the time is long, but if we use this method, it can reduce the amount of data, improve the efficiency.