利用Hausdorff距离的快速人耳检测

为有效解决人耳检测中的遮挡和光照不均等问题,结合Hausdorff距离的模糊匹配的特点和较强的抗干扰能力,提出基于Hausdorff距离的人耳检测算法.算法首先用肤色模型检测出人脸区域,然后用canny算子提取边缘,并进行距离变换,再构造多分辨率金字塔模型,在不同分辨率层上采用不同的匹配策略进行Hausdorff距离匹配.算法采用多分辨率方法显著提高检测速度,采用部分Hausdorff距离提高它的抗干扰能力.仿真实验结果取得95%的正确率,并且对遮挡和光照不均具有较强的鲁棒性.

Fast Human Ear Detection Using Hausdorff Distance

For the purpose of resolving the problems of occlusion and poor illumination on the ear image,the human ear detection algorithm based on Hausdorff distance is proposed combining with the feature of fuzzy matching and the strong anti-jamming ability of Hausdorff distance.Firstly the algorithm detects the region of face by using face color model.Secondly edge is detected by canny operator,and then the distance transform is computed.Thirdly the multi-resolution pyramid model is constructed.Lastly the image is matched by Hausdorff distance whose different method is used on the different resolution layer.The algorithm uses multi-resolution method to improve detection speed,and uses partial Hausdorff distance to increase its anti-jamming ability.The experiment result shows that the correctness is 95%,and has strong robustress for occlusion and poor illumination.