| dc.description.abstract | The purpose of this research is to improve the effect of shadows on Shape From Focus(SFF) with the exponential correction method. The SFF is mainly used to measure the
surface topography and three-dimensional reconstruction of objects, which can measure objects in full field. In the process of machine vision, external environmental factors, such as lighting, vibration, etc., often lead to poor image quality. This paper will improve the impact
of shadows.
The principle of SFF is to place the object to be measured on a reference plane, move the CCD or the object to be measured by a known distance, and continuously capture images during the displacement process, judge the focus based on the clarity, and then calculate the depth map. It is mainly based on the difference in light intensity in the neighborhood of each image to determine the value of focus of its surface topography. The value of focus is calculated by the focus algorithm. In this study, we found through experiments that shadows on the surface will reduce the light intensity of the image, resulting in a smooth surface or lack of texture. As a result, the system cannot distinguish the difference between each image. Therefore, before using the focus algorithm to calculate the image, the image is corrected. We use a variety of algorithms to calculate the focus value, and experiment with or without shadow on the surface. We use the exponential correction proposed by this research to correct the image and analyze the difference before and after correction. The measurement performance of this system is verified by a white light interferometer. The spatial resolution of the system is 1 um/pixel, and the axial resolution is between 50 ~ 100 um, and the experiment correction can reduce the error by at least 10 %. | en_US |