基于Radon变换的健美操跳跃动作轨迹实时提取方法

Real Time Extraction Method of Aerobics Jump Trajectory Based on Radon Transform

在健美操跳跃时若不能及时明确指出不标准跳跃动作,运动员极易受伤且得分较低,因此,提出基于Radon变换的健美操跳跃动作轨迹实时提取方法。采用成像技术完成健美操跳跃动作轨迹图像采集;利用Radon变换将采集图像初步去噪,并利用图像分割、小波多尺度分解方式把图像分成多个子模块去噪,完成图像去噪;运用支持向量机方式得出去噪后图像子模块特征分布情况,并利用视觉RGB方式分离特征,凭借虚拟视景重构方式将子模块特征融合,输出结果。实验结果表明,所提方法提取结果精度、拟合优度均较高,与实际情况基本一致,并且提取速度极快,具有良好鲁棒性。

If aerobics can’t clearly identify non-standard jumping actions in time, athletes are easily injured and have low scores. Therefore,a real-time extraction method of aerobics jumping movement trajectory based on Radon transformation is proposed. It uses imaging technology to complete the image acquisition of aerobics jumping motion trajectory;uses Radon transform to initially denoise the captured image, and uses image segmentation and wavelet multi-scale decomposition to divide the image into multiple submodules to complete image denoising;uses Support Vector Machine method obtains the feature distribution of the image submodules after denoising, and uses the visual RGB method to separate the features, and uses the virtual view reconstruction method to fuse the sub-module features and output the result. The experimental results show that the extraction results of the proposed method have high accuracy and goodness of fit, which are basically consistent with the actual situation, and the extraction speed is extremely fast and has good robustness.