Research On Image Feature Point Matching Algorithm For Visual Indoor Positioning

With the improvement of science and the improvement of people’s living level, the location based service and attracts more people’s attention. In the indoor environment, the environment is relatively complex. Taking into account the effect of the house walls, objects, mutual occlusion and shielding, personnel in the indoor environment can move freely and other factors, GPS system and the traditional outdoor positioning system does not show ideal effect. Now the vis ual-based positioning technology is popular because of its simple equipment, less affected by the factors that have been widely concerned. Due to the camera mobile ph one has become the standard belongings, so visual positioning no need to add additional equipment. Also due to the changement of construction is small, the affected factors of visual based positioning are smaller.Image matching is the most critical technology in the visual indoor location system. In this system, to get the position of the users is to obtain the image and offline database image for matching. It can be seen that the image matching speed, accuracy and robustness are direct impacts on the localization speed, accuracy and robustness. This paper focuses image feature points matching algorithm on the visual system for indoor localization.Firstly, the paper expounds the visual indoor positioning technology and image feature point matching technique, and the domestic and abroad research status of eliminating the matching error technology. Studying the theoretical knowledge of epipolar geometry location method, and building the indoor positioning system based on visual.(1) Aiming at the problems of low efficiency in the traditional visual indoor positioning algorithm by SURF algorithm. This paper using the AGAST corner point detection algorithm and FREAK binary description algorithm for image feature extraction, in order to accelerate the speed of image feature extraction and matching, at the same time, guarantee the accuracy of image matching.(2) Aiming at the problem of slow and lack of stability of traditional RANSAC algorithm, this paper proposes the improved RANSAC algorithm. By giving matching quality to the corresponding matching paris, selecting the matching pairs of quality to calculate the projection matrix. Reducing the number of iterations, in order to increase the speed and ensure the stability of RANSAC algorithm by setting adaptive parameter.Finally, this paper established visual indoor positioning system on MATLAB, and introducing the process of using the system. The original algorithm and the proposed algorithm are analyzed and compared by simulation on this system. Simulation results show that the algorithm proposed in this paper can gua rantee the positioning accuracy, while getting a substantial increase in the efficiency of localization.