| With the rapidly development of economy and technology, informationization and digitization have gained high popularity among the people, indoor localization technology is developing towards general and intelligent mode. On basis of analyzed the traditional algorithm, this paper propose an indoor localization algorithm based on image features detection in order to realize convenience and accuracy. It is essential to confirm the image transform relationship by utilizing image features. And The difficulties to be resolved in the method is how to extract stable feature points and build feature descriptors that are adaptive to transform, distortion, faintness, noise and factors in other forms.This paper introduced the several popular approaches of tracking in detail, and focused on the study of indoor localization based on image feature matching. On the basis of analyzing and summarizing the approaches to feature point extraction, this paper presents a fast image matching algorithm based on SIFT corner detection. The algorithm is simplified by analyzing the characteristic of floor image, so it reducing the sample octave and the operation complication of Gaussian convolution with the quantity and stability promised by selecting appropriate parameters. The invariant feature points are produced by Harris algorithm at each octave. Following of this, a descriptor which is 128 dimensions is structured based on these points, and false matches are removed by the method of Random Sample Consensus. Therefore, two images can be matched through this way accurately. After that, this algorithm is used in indoor localization and validated by emulator experiment at aspects of scalar, rapidity and environment adaptability. The results show that our algorithm almost performs as well as SIFT in the fact of matching precision, but the process speed is about 30% more than it. At the same time, the precision of our algorithm is about as twice as the tradition indoor localization algorithm. |
PDF Full Text Request