实时人耳跟踪作为人耳识别系统的前提，是一个意义重大且具有挑战性的课题。为了实现对人耳的快速实时跟踪定位，提出了一种融合肤色及灰度轮廓信息检测跟踪序列图像中人耳的新方法。该方法首先采用改进的CAMSHIFT算法在肤色概率分布基础上，初步对侧面人脸这一包含耳朵的大面积肤色区域进行粗定位，进而根据耳朵自身丰富的轮廓信息，应用轮廓提取及拟合来进一步精确定位人耳。实际应用表明，该方法具有速度快、效率高和鲁棒性好的特点，不仅能够满足实时性的要求，并且对于干扰和耳朵自身一定角度的偏转都有较好的容错性。此外，该方法在相对复杂的背景下仍能获得较好的跟踪效果。

Robust real-time human ear tracking is a challenging task. In order to realize the fast tracking and locating of the human ear,a new method is proposed that combines both skin-color information and intensity contour information to detect and track the human ear in the sequence frames. CAMSHIFT as a fast computationally efficient algorithm on color-based tracking is applied to roughly track the side view of human face as the region of interest(ROI) and then contour fitting is operated in ROI,which can improve the calculation speed obviously,for further accurate locating due to the ample contour information the ear contains. Experimental results show that this method is fast,robust and effective,which can meet the real-time requirement and has certain tolerance of distractors and angle change of the ear caused by the head rolling. Furthermore,the method can be implemented under relative complex background situation and works well in practice.