Abstract: Corner detection is widely used in real-time systems under complex environments,e.g.,the augmented reality system,and is crucial to implement the tracking registration.The classical fast corner detection algorithm performs poorly under noise and strong light conditions,while Harris corner detection algorithm cannot work in real time.Therefore,both of them cannot well satisfy the requirement of the augmented reality system.To address these problems,a novel corner detection algorithm,the star-based grayscale projection,is proposed in this paper:the star projection value of all pixels in the detection area is computed on the basis of extracting the edge of the image,and the false corner points are eliminated gradually by judging the project value of the main peak area and the distance of the main peak.Experimental results show that the detection results of the proposed algorithm are compared to the existing algorithms in two aspects of real time and robustness.The proposed method is robust to noise and strong light,hence it is suitable for real-time systems under complex environment.