数字图像中的拐点通常成为重要的信息载体，因此准确、稳定和实时地检测出拐点便成为拐点检测算法面l临的主要问题，针对该问题，提出了一种新的二值图像中拐点的实时检测算法。该算法与传统基于边界链码的拐点检测算法不同，其是首先构建像素的k(k>8)邻域，并将图像中物体的边界表示为k邻域链码；然后根据曲率定义的差分形式计算各边界点处的曲率；最后通过检测曲率直方图的局部峰值精确定位出拐点，并利用拐角内部像素的颜色统计信息迅速判断出拐点的凸凹性。(The inflection point in the digital image are important information carriers, so accurate, stable and real-time detection of the inflection point will become the main problem of inflection point detection algorithm face L, to solve this problem, we propose a new two value image in the inflection point of the real-time detection algorithm. The algorithm with the traditional inflection point algorithm of edge detection based on chain code, which is firstly constructed pixel K (k>8) neighborhood, and the image object boundary representation for K neighborhood chain code; then according to the definition of curvature curvature difference form to calculate the boundary point; precise local peak location finally tested by the curvature histogram of a turning point, and the use of internal corner pixel color statistical information to quickly determine the convexity of the inflection point.)