Research On Mobile Robot Localization Based On SURF

Mobile robots are widely used in modern society because of their features of flexibility in action and ability to perform complex work. Autonomous navigation function is the most basis and important function that mobile robots must have, and it is concerned by researchers of many different fields. The SLAM technology is the most used technology when mobile robots are navigated in the unknown environment. The vSLAM system gets the positions of mobile robots using visual sensors. The images captured by visual sensors contain relatively abundant information, and some special information can be recognized. So the vSLAM system plays a important role in the implementation of SLAM technology. However, there still are some problems in the vSLAM system, such as relatively complicated visual information, relatively high computational complexity and the difficulty to get the big resolution image from visual sensor in the same time. Therefore, the concerned problem by the researchers in this field is improving the speed of formation and match between detector and descriptor based on the vision of the localization system.Background knowledges of robot navigation was first presented in this paper, reviewed common operators used in localization realization of vSLAM system, and introduced the realization bases in the localization part of vSLAM system-local invariant features theoretical basis and mathematical reasoning in more details.Then explained the principle and procedure of the most brilliant algorithms SIFT and SURF deeply. According to the features of vSLAM system, to improve the speed optimize the process that SURF algorithm generates descriptors.Finally, this paper reviewed image match theoretical basis and mathematical reasoning, introduced quick index in the SURF matches, and proposed SURF point matching algorithm based on region growing. This algorithm can maintain matching reliability and extremely improve matching speed.