An improvement algorithm on RANSAC for image-based indoor localization

With the progress of science and technology, mobile phone become a necessary carryon in our daily life. Location service get more and more people's attention. But in the indoor environment, due to the block of walls and other factors, traditional positioning systems are often not available such as GPS. Now the visual-based indoor localization have become a popular research in recent years. We proposed an improvement of the original RANSAC method and apply it in the indoor visual-based localization. In our improving proposed method, we define a matching quality function to measure the matching quality of a matching pair of images. Through matching quality, we choose the 4 best matching pair to calculate the projective transform matrix of two images and eliminate false matching pairs. The most important improvement of proposed algorithm is sorting the matching pairs according to matching quality instead of random testing. The experiment result shows that the proposed method significantly reduce the iterations and running time, at the same time, improves the stability of the algorithm.