基于SUSAN原理的黄河模型角点检测

摘要

This paper analyzes the principle and availability of SUSAN algorithm in comer detecting. Based on this, a method that adopts the improved SUSAN algorithm is put forward to extract the comer features of the Yellow River Model images. The method utili- zes first the self-adaptive threshold method of gray level difference, and then adopts the method of setting its upper and lower limits of geometric threshold to realize the automatic identification and detection for the Yellow River comer. Experiments show that this method of comer detection is good for de-noising. Its boundary is clear, true, meticulous, and the location is accurate. It makes the foundation for detecting stereo matching and 3D reconstruction in next step.%分析了SUSAN算法进行角点检测的原理和有效性，在此基础上提出了一种采用改进了的SUSAN算法来提取黄河模型图像的角点特征。利用自适应的选取灰度差阈值的方法；再采用利用设定几何阈值的上下限方法，实现了黄河模型角点的自动识别和检测。实验表明，该方法提取的角点抗噪性能好，清晰真实，细致，定位精确。为下一步立体匹配和三维重构打下了基础。