Design And Implementation Of Static Important Objects Detection For Smart Environment Explorer Stick

The World Health organization (WHO) estimated that in2013,285million ofthe world’s total population was visually impaired. Visually impaired people facemuch inconvenience when interacting with their surrounding environments, and themost common challenge is to indoor navigation. While many literature and systemshave been focusing on navigation, way-finding, text reading and so on, there are rarecamera and sensors based system available in the market to help the visuallyimpaired people for the indoor space navigation, and integrate orientation andmobility functions with economic efficiency.The Smart Environment Explorer Stick (SEE-Stick) has been proposed in thispaper and aims at implementing a real-time, low-cost and energy efficient smartstick providing VIP with both navigation and environment detection functions.Considering the importance of reliability for VIP and low cost, multiple sensors datafusion method has been introduced to improve the navigation accuracy. It comprisesa MEMS grade gyroscope, a wheel encoder, a consumer grade GPS receiver, and asingle camera. With the single camera, several methods have been developed andtested on the real system, such as texture based segmentation, shape based objectdetection, and machine learning based object detection. The final method combinesall the advantages of these method to produce an accurate result, at the same time,considering the platform resources limitation. At a specific location, an image istaken from the camera, fusing with the data from other sensors, it is able torecognize the static important objects, which will help the whole system navigate theVIP in an indoor space. The algorithm utilizes the texture based method to filter therough objects, and then segment the area, and detect the objects with pre-trainedSVM vector.The object detection module demonstrated the predicted result of SEE-Stick forobject detection, and provided important information for the indoor navigation. Theresults validated the efficiency, reliability and accuracy of the method developed inthe object detection module. And also, the module is integrated with other modules,which work together to facilitate the VIP interact with the unfamiliar indoor space.