复杂背景及复杂运动前景下的指尖检测方法

人手指尖空间位置的检测是基于视觉的人机交互中的关键问题,该文提出了一种在复杂背景以及复杂运动前景的情况下也能快速、准确、鲁棒得检测指尖的方法。针对传统方法必须对背景、前景设定约束的问题,该文提出了感兴趣运动前景提取算法,即仅提取包含运动人手的前景图像。针对传统方法造成指尖误检测的问题,提出了结合多种人手轮廓关键几何特征来有效得检测指尖的算法,当采集的图像中未包含有伸出的手指时,该算法也能及时作出无指尖判断。实验结果表明该方法对指尖位置检测及无指尖的判断均有较高准确率。

Fingertip Detection Approach from Cluttered Background and Moving Foreground

Fingertip detection is one of the key technologies in vision-base human-computer interaction.The paper develops a fast,accurate and robust method of fingertip detection in cluttered background and moving foreground.For the problem of the traditional method must set constraint on background and foreground,the paper presents an algorithm of interested moving foreground detection,which only extract the image with hand foreground.For the problem of the traditional method makes mistakes in fingertip detection,the paper presents an effective algorithm to combine lots of key geometric features of hand＇s contour to detect fingertip.If there is no outstretched finger is in the images,the algorithm can output the correct judgments of no fingertip.Experiments show that the algorithm can effectively detect fingertip.