UAV image mosaic based on adaptive SIFT algorithm

Abstract

UAV has many advantages such as flexibility and security, and is becoming an important supplementary method of remote sensing. However, the instability of UAV platforms usually results in poor image quality, such as big rotation angle, leading to difficulty in creating image mosaics with conventional methods. The SIFT algorithm is nowadays widely applied in image mosaics, but it also has many disadvantages. One of which is that the distance ratio threshold of SIFT is usually defined as a fixed value, which determines the number of matching points. In this paper, an adaptive method of determining distance ratio threshold is proposed. To determine whether a threshold is the optimum value, the ratio of RANSAC optimized points and matching points is used as an evaluating criterion to identify the threshold performance. Experiments show that the adaptive method can improve the matching accuracy while not bringing too much calculational burden. Image mosaic results are also good with the adaptive SIFT method.