Key Technology Research Of Improved Slam Based On Image Feature Point Matching Algorithm

With the rapid development of science and technology, mobile robots become more and more intelligent and autonomous. Simultaneous localization and mapping (SLAM) is a fundamental issue and a hot topic in the field of robot navigation. The so-called map building is an accurate description of the positions of every object in the environment. The simultaneous localization refers to determining own position in the environment and updating map. Compared with ultrasonic and laser sensor, visual sensor can get more information and reduce the running time, and then mobile robots become more easily adapt to the environment. This paper studies image feature points for mobile robot on the basis of binocular vision.First, in this paper, the motion and observation models of mobile robot are given. It introduces the odometer models and camera models, elaborates the principle of binocular vision. Getting the position coordinates of the environmental characteristics by camera calibration and coordinates transformation, then applying it to the follow SLAM studies. Second, in view of ORB algorithm does not have scale invariant, the ORB algorithm is improved combined with SURF algorithm named SURB algorithm. The feature points of SURF are extracted and then builds ORB descriptor. During feature matching, splits the matching image into two parts, which reduces image search time. Then, hamming distance completes the coarse matching, which avoids global feature points extraction and matching. Finally, false matching points are removed by PROSAC algorithm to improve matching accuracy. Third, cubature Kalman filter selects a set of points according to the volume ruler, gets statistical characteristics of random variables by passing into the non-linear function. Compared with the extended Kalman filter and unscented Kalman filter, the state estimation errors of CKF-SLAM are greatly reduced. However, the position of CKF-SLAM becomes ineffective when the system changes. In order to solve this problem, this paper proposed a new solution named strong tracking square-root cubature Kalman filter based on SLAM algorithm(STF-SRCKF-SLAM). Strong tracking filter is the main theoretical basis of STF-SRCKF, it can ensure the model robust even with uncertain system and it can avoid possible filtering divergence while preserving the SRCKF value of the stable characteristics.Last, it can be found that improved algorithm helps mobile robot be more familiar with environment, more faster and get higher accuracy in targeting, more flexible in the environment by several times experimental simulations.