Corner detection in planar curves is a basic problem in image processing domain. The authors start from concluding corner's intuitive characters, introduce fuzzy set theory, and propose a multi-feature based algorithm which contains fuzzy subjection degree to achieve the goal. First, two simple fuzzy characters are addressed which compile with human perception. Second, three formulations corresponding to corner features are presented, and they are synthesized and summarized into a principle used to detect and locate corners. At last, by comparing computed results with existed results, robustness and effectiveness of our algorithm are proved.