【作者】 史萌；

【导师】 来跃深；

【作者基本信息】 西安工业大学 ， 机械制造及其自动化， 2012， 硕士

【摘要】 随着图像处理技术与计算机技术及检测技术的结合,形成的视觉检测系统可以完成对微小尺寸的高精度测量,是其它设备所无法完成的。而自动聚焦是视觉检测系统中的一项关键,是图像处理、模式识别、精密定位等后续操作的基础。只有快速、精确地采集到清晰的图片,才可以有效地完成后续检测工作。本课题以刀具检测仪为研究平台,围绕着自动聚焦技术研究了图像的清晰度评价函数、焦点搜索算法和聚焦区域选择等相关技术,得出适用于刀具图像的函数算法。主要包括以下几个方面：1)了解课题的来源和背景,掌握自动聚焦技术的国内外发展状况和关键技术,并对与自动聚焦技术相关的理论原理作了详细的解释说明。2)分析了自动聚焦算法的相关理论,包括三个模块：清晰度评价函数、聚焦搜索策略算法和聚焦区域选择的影响。清晰度评价函数是对刀具图像的离焦程度进行评价,它是有参考的图像质量评价,本文对常用清晰度评价函数做了详细的分析比较,研究结果表明8邻域拉普拉斯算子在处理刀具图像时精度高；聚焦搜索策略算法的目的是实现聚焦点的搜索和定位,快速找到精确聚焦位置。本文分析了常用的焦点搜索算法,对算法的流程进行了说明,结果表明对爬山算法改进后的“三点判别法”避免了局部峰值造成的误判,精度高；聚焦区域选择是选择感兴趣区域进行聚焦,可以减少计算量。本文分析了各区域选择算法的优缺点,将其与整幅图像的调焦曲线进行对比分析,结果表明多点取窗区域选择算法效果好,计算效率高,避免了对整幅图像进行计算时的数据量大、时间长等问题。更多还原

【Abstract】 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.更多还原