마커 없는 증강 현실 응용을 위한 조명과 시점 변화에 강인한 코너 검출

Corner detection is the first step of many vision tasks such as matching, tracking and recognition. In this paper we proposed a method to improve the existing FAST corner detector so that increased the stability of markerless AR. The repeatability of the corner detector is increased by using adaptive threshold. We use only repeatable corners for the registration of a reference image to improve the matching accuracy. Experimental results show that the proposed corner detection gives higher recall-precision ratio than the previous FAST corner detection method.