乒乓球正手拉球动作图像正确姿势轮廓提取方法

Extraction Method of Correct Posture Contour of Forehand Action Image of Table Tennis

由于图像重合度较差造成提取结果精度较低的问题,设计乒乓球运动员正手拉球动作图像正确姿势轮廓提取方法。采用视觉系统完成其正确姿势图像的获取,设定方向向量提升获取结果的有效性,使用卷积神经网络对获取到的图像进行处理,提高图像的重合度。使用边缘图像轮廓提取算法对处理后的图像进行轮廓提取,完成乒乓球运动员正手拉球动作图像正确姿势轮廓提取。实验结果表明,所提方法图像重合度高于三种传统方法,重合率区间相对较小,提取精度更高。说明所提方法的图像轮廓提取效果较好。

The poor coincidence degree of the image lows the accuracy of the extraction results.Therefore,extraction method of correct posture contour of the table tennis players'forehand pulling action image is designed.The vision system was used to get the correct pose image and set the direction vector to improve the effectiveness of the results,and to use convolution neural network to process the acquired image and improve the coincidence degree of the image.The edge image contour extraction algorithm was used to extract the processed image.The correct position contour extraction of the table tennis players'forward hand pulling action image was completed.The results showed that the coincidence degree of the proposed method was higher than that of the three traditional methods.Moreover,the coincidence rate interval was relatively small and the extraction accuracy was higher.It proved that the proposed method has a good effect on image contour extraction.