Abstract
In this study, a novel 3D non-contact microscope system is developed. In this system, a CCD camera and microscope system are used to grab object images. A stepping motor is used to move the object vertically. In this study, the technology of Shape-from-Focus is adopted to estimate the 3D geometry of examined object. The images have a status of the shift that is due to the optic axis and mobility platform are not vertical. We make use of position compensation to correct the shift of the images. Sum-Modified-Laplacian(SML) and Sum-Enhanced-Modified- Laplacian(SEML) operator are generally used in this approach to identify object in focus. After an operation of ML and EML, we join Gaussian filter to remove noise in the images. Median filter can filter out the isolated points and preserve the edge contour of the images. Gaussian fitting and interpolation can eliminate noise and show the height of the objects. We rebuild the contour and define the index to optimize the method to reach a better result. At last, we adopt the best method to measure the height of the solderballs.