一种基于机器视觉的引体向上计数方法

针对引体向上项目人工计数效率低、误判率高等问题,提出了一种基于机器视觉的引体向上计数方法.利用AdaBoost算法构建Haar-like特征人脸人手分类器,通过计算图像不变矩获取人手人脸面积和质心坐标,分别取人脸面积、人脸与人手质心垂直距离做阈值;利用混合高斯背景模型提取运动前景,统计图像序列中ROI区域灰度变化,然后做椭圆肤色检测,计算图像不变矩获取图像序列中人脸质心坐标,通过对每帧的灰度值和质心坐标变化做分析,并与设定阈值作比较,得到引体向上数目.实验表明,该方法能快速的对复杂环境下引体向上动作进行计数,计数准确率约为91.3％.

A method for pull-up counting based on machine vision

Aiming at the problem of low efficiency and high probability of misjudging in the pull-up project,a method for counting of pull-up which based on machine vision technology is put forward in this paper.The Adaboost algorithm is used to construct a face and hand classifier of Haar-like feature.The human face area and the centroid coordinate value are obtained by computing the invariant moment of the image,the face area and vertical distance of face and hand are taken as threshold respectively.The Gaussian mixture background model is used to extract foreground,the gray scale changes of ROI areas is count in image sequence,and elliptical skin detection is used,the centroid coordinate is obtained by calculating the invariant moments of image,the gray value and the centroid coordinate of each frame are analyzed,and then comparing with the corresponding threshold,and the number of pull-up will be obtained.The experimental results show that the method can count pull-up action in the complex environment quickly and the accuracy rate is about 91.3％.