Research on the Outdoor Practice Model of College Sports Using Multiple Mixed Information Technology

In order to improve the feature recognition ability of sports video deep model transfer learning, this project intends to study a sports image classification algorithm that integrates constrained Boltzmann machine and convolutional neural network to enhance the recognition ability of sports images. Firstly, this paper proposes a structured topic classification algorithm for sports video images. Then a convolutional network based on iteration of constrained Boltzmann machine is designed to transform the existing convolutional neural network into a small target. The maximum levels of constrained Boltzmann computers and software are retrained. BP neural network algorithm was used to modify the parameters of the model. The limited Boltzmann machine hierarchy proposed in this project can not only realize the complete connectivity of all feature quantities, but also extract the set of objects with specific characteristics from the maximum log-likelihood theory. The experimental results show that the algorithm has better anti-noise, wider sports video detection and tracking performance.