A Method for Standardizing Motion Segmentation of Teaching Video Images for Improving the Quality of Sports Training
In order to improve the training quality of sports events,a method of standardized action segmenta-tion of teaching video images was designed.Since gray level image only contained brightness information,but with-out color information,the gray level conversion of teaching video image was implemented to improve the processing speed of image and reduce the subsequent computational complexity.The feature transformation was implemented based on the spatial gradient feature,and the method of combining the gradient feature and Gaussian convolution operation was used to extract the action feature in the time-domain,and the action was quickly detected by the sta-tistical anteroor-anteroor-interframe difference projection feature of the action sequence.The multilevel sequential convolutional network was improved,and the performance of the traditional sequential action segmentation model was improved by adding a cascade module,a double expansion layer and a boundary generation module.The test results showed that the standard action segmentation effect of the design method had clear edges and complete details,and the background elimination effect was better.With the increase of movement complexity,the editing scores and F1 scores of the design method only decreased slightly,and the overall performance was relatively stable and good.
sports training qualitygrayscale conversionteaching video imagesnormative action segmenta-tionphase cascade moduleboundary generation module
NETL
NSTL
万方数据
