Basketball player detection technology is an important subject in the field of computer vision and the basis of related image processing research. This study uses machine learning technology to build a basketball sport feature recognition model. Moreover, this research mainly takes the characteristic information of basketball in the state of basketball goals as the starting point and compares and analyzes the detection methods by detecting the targets in the environment. By comprehensively considering the advantages and disadvantages of various methods, a method suitable for the subject is proposed, namely, a fast skeleton extraction and model segmentation method. The fitting effect of this method, whether in terms of compactness or quantity, has greater advantages than traditional bounding boxes, and realizes the construction of dynamic ellipsoidal bounding boxes in a moving state. In addition, this study designs a controlled trial to verify the analysis of this research model. The research results show that the model proposed in this paper has certain effects and can improve practical guidance for competitions and basketball players training.