위치 추정을 위한 영상 특징점 정보 검출 방법

Abstract

This paper investigates a method to find coordinate of a feature captured by the image sensing device Kinect provided by microsoft. It detects landmarks from the image and calculates their coordinates. From the color image and depth information by the Kinect, it extracts features from the background. It uses channel transformation, binary transformation, labeling, convex hull algorithm for the extraction. Once feature is extracted, it's coordinate is calculated. For these processes, it uses OpenCV library. To remove the discrepancy between the color image and depth image, coordinate adjustment is needed. The coordinate is represented with respect to the Kinect coordinate frame. The method is tested and verified indoors. The proposed method can be used for localization, mapping, and simultaneous localization and mapping(SLAM) which are indispensable for autonomous navigation of mobile robots.