A matching method is proposed for zero-insulator detection，which combines the Scale Invariant Feature Transform (SIFT)algorithm and an improved random sample consensus (RANSAC) algorithm. Firstly, SIFT algorithm and Euclidean distance function is adopted to extract image features and pre-matching between the testing zero-insulator string image and the standard zero-insulator string image in the image library. The Euclidean distance may cause mismatches of some features but Mahalanobis distance algorithm measures the distance between the data by covariance distance and the correlation between the components of the eigenvector is taken into accounted. So, the mismatching features are wiped out by RANSAC method combined with Mahalanobis distance algorithm in this paper. For it takes too much time by RANSAC method on the data and the models’ validity tests through all the data. An adaptive method based on the ‘outers’ statistic rate of sampled digital models is presented to minimize the running time for the data and models’ validity tests. Results of the experiments shows that the proposed method achieves precise and prompt detection for the zero-insulator.
