표지판 인식을 위한 2D 및 3D 정보 융합 기술

Abstract

ABSTRACT

Information convergence technologies between 2D and 3D space for a sign recognition

Kyong-Jo Hwang Advisor : Prof. Sang-Woong Lee Ph. D. Department of computer Engineering Graduate School of Chosun University

Most of the existed researches about sign detection and recognition are based on color, shape and the combination of color with shape. Usually, there are two major steps about the sign recognition. the first step is to detect the sign and the second one is to recognize the sign form the extracted information. In order to recognize the sign, the sign detection should be performed first. However, the conventional method only use one camera to capture the 2D information about the sign which can not provide enough information for the sign detection and recognition. Therefore, it is promising to use the 2D information and the distance information which can provide more detail about the sign for the detection and recognition.the combination of 2D and 3D image can achieve good recognition result. In this paper, method based on 2D and 3D information is proposed to detect the signs. This method uses a stereo and kinect camera images to acquire 2D image and 3D depth. The distance information is extracted from the 3D depth image and then the unnecessary range is removed by the obtained distance information using filtering approach so as to measure the effective distance of the target. Then, the region segmentation method is applied to the effective distance image and the individual ROI (region of interest) can be identified. The extracted ROI can also be converted to HSI image to facilitate the sign detection. Experimental results show that this method has a good performance in terms of sign detection.