田文利(陕西财经职业技术学院, 咸阳 712000)

基于图像清晰度评价函数与自动取窗对焦的目标视觉调焦算法

田文利
(陕西财经职业技术学院, 咸阳 712000)

为了解决当前显微视觉系统在视场小、景深短和聚焦精度要求高的状况下存在离焦问题，难以满足精密制造业的高精度测量的要求，分别从清晰度评价、聚焦窗口动态选择和对焦搜索的角度出发，提出了基于清晰度评价函数与自动取窗对焦的工业显微视觉调焦系统。根据灰度梯度函数和频域分析函数，设计了耦合空间域与频域的清晰度评价算法，全面评价图像清晰度。对图像进行分块处理，根据子块灰度梯度变化程度，分离感兴趣区域与背景，实现动态智能选择聚焦窗口的目的。根据像方焦平面的单峰性特征，动态调整多步长与单步长，改进了爬山搜索算法，准确定位焦平面，提高聚焦准确度。实验测试结果显示：与当前显微视觉调焦技术相比，本文算法具有更高的准确性与实时性。

显微视觉； 清晰度评价； 聚焦窗口； 爬山搜索； 灰度梯度函数； 对焦搜索

Abstract： In order to solve the current micro vision systems such as small visual field, short depth and insufficiency from coke in high accuracy requirement, from the definition evaluation algorithm, dynamic focusing window selection and focus search algorithm, this paper presents evaluation function and the automatic fetching window for industrial microscopic visual focusing system. First of all, according to the gray gradient function and frequency domain analysis function, it designs the coupling of clarity evaluation algorithm of time domain and frequency domain. And then, it blocks the image, based on the sub-block gray gradient change degree, to distinguish the interested region and background, and achieves the goal of dynamic intelligent choice focused window. Finally, according to the unimodal feature of the focal plane, like the organic adjustment of step length and step length, it puts forward an improved climbing search algorithm. Test results show that compared with the current microscopic visual focus systems, this system has higher accuracy and better real-time performance.

Keywords： Micro vision; Sharpness evaluation; Focusing window; Hill-climbing search; Gray level gradient function; Focus search

0 引言

精密制造行业对于零部件的测量要求往往很高，需要用到显微视觉系统，对极小的物体进行成像测量，达到精密测量的目的[1-2]。但是显微视觉系统由于本身的局限，成像视野小、景深小，从而对系统的快速准确聚焦能力很高，从而限制了精密制造业的进一步发展。

在显微视觉方面，国内研究人员已经取得了一定研究成果……