In order to improve the special training effect of sports movements, this paper combines machine vision image coarse-grained technology to build a special training system to improve the special training effect of sports. By building an image acquisition model, the movement area of special action is extracted, and the recognition action is separated from the background, so as to improve the recognition effect. By setting the image threshold, the target recognition action is segmented from the background image, which effectively improves the accuracy of subsequent recognition. Moreover, this paper collects the images of athletes’ special actions, and obtains the structural data and edge features of special actions by integrating action information database, equipment model and other models. In addition, this paper scans the features and matches them with the action identification information, so as to realize the image acquisition of the recognized object. Finally, through the experimental results, we can see that this system has a good effect on the recognition of wrong technical movements, which can effectively improve the special training effect of sports movements.