No abstract

This paper presents an approach for identification and reconstruction of surfaces from unorganized three-dimensional data. The concepts of distance function and free differential motion are introduced to characterize the kinematic properties of some kinds of surfaces, including simple surfaces and kinematic surfaces generated by a generatrix undergoing a screw motion. Then, a method is proposed automatically to identify the type of the surface and roughly estimate the location of the screw axis as well as the profile of the generatrix from the discrete measurement points. Based on the differential property of the signed distance function, a least-squares surface fitting algorithm is developed to finely tune the location and shape parameters of the reconstructed surface. This algorithm is also applicable to the reconstruction of implicit surfaces such as quadratic surfaces. Examples are provided to investigate the efficiency and precision of the developed method.

As a common kind of surface defects on multi-crystalline silicon solar cells (about 65% of total defects). Based on HALCON image processing library, an automatic silicon wafer surface spot defects detection and classification system has been developed: captured color images via CCD image sensor, located wafers and spots, extracted the spot features, computed and separated the qualified regions from the unqualified ones by SVM classifier. The experimental result showed a high accuracy of 95.7% and fast image analyzing and classifying process (less than 600ms).