Studies on Image Stitching Algorithms in Machine Vision Inspection of Solar Panel

Abstract

Image stitching is an important technology in machine vision inspection whose algorithms focus on the detection and match of feature points. For the solar panel images achieved by machine vision, it’s found that the available image stitching algorithms failed to detect enough valid feature points, which led to a large number of incorrect matching points. Here, we improve an algorithm to locate feature points and find dense match points. The proposed algorithm is based on SIFT. The improved algorithm (I-SIFT) can locate the position of feature points in every solar panel images, reducing the effect of invalid feature points on the experiment. Euclidean distance is used to preliminarily ensure the matching points, and then the relative horizontal position of every matching points is used to eliminate the mismatching points caused by the space similarity of the feature points, so that the matching accuracy is improved. Then the improved algorithm is used in the traditional Harris, SIFT and SURF stitching algorithm. The experimental results show that the success rate of I-SIFT algorithm can exceed 95% and the computation time has decreased by nearly 95% than that of the traditional algorithm. In conclusion, the improved algorithm implements accurately and rapidly stitching for solar panel images.