Wearable Visual Aided Positioning System

Blindness and visual impairment have been one of the serious public health problems. It is very difficult for the blind and vision impairment patients to safely walk in the unknown environment. Nowadays most of the existing products which are used for helping blind people to walk are based on the GPS positioning system. These products have many inherent defects, thus it needs to further research a better positioning system that is more robustness, highly precision, and widely used to help. The stereo vision localization algorithm has many advantages, its three-dimensional positioning results are quite precise, and thus it can be used for the auxiliary positioning system.In this paper, we design a wearable visual aided positioning system that can be carried on the user’s back. While the user is walking, the binocular camera continuously takes pictures about the surrounding. Then the system can provide the user with real-time location information according to the principle of stereo vision.This paper studied the research status of the wearable visual positioning system and the mobile robot localization algorithm, and then analyzed the technology methods used in the existing system and their advantages and disadvantages in detail. After that we designed a wearable vision positioning system, which integrates the stereo vision localization algorithm to provide real-time position information.Feature tracking is an important step of the stereo vision localization. In the wearable localization system, the camera will shake violently during the execution of positioning, and the conventional localization algorithm may fail. This paper proposes a robust stereo localization algorithm that integrates the inertial aided KLT feature tracking algorithm, and uses this algorithm to solve the above problems in the practical application. The inertial aided KLT feature tracking algorithm can narrow the searching area and increase the probability of convergence, thus it can improve the accuracy of the positioning results. Through the combination of the theoretical analysis and experimental methods, we achieved the calibration of the camera and the IMU as well as the information synchronization issue. By the results of the experiments in the real environment, the performance of the localization algorithm we proposed and the effectiveness of the system in this paper are well validated.