I. Introduction
Athletes' injuries affect the athletes' physical and mental health to varying degrees, which is harmful to the improvement of athletes' sports performance. In the process of football training, the probability of getting knee injury is gradually increasing, which has always been one of the factors that influence college football training [1]. The combination of artificial neural network with college football training offers a method of using artificial neural network to strengthen football training, and uses artificial neural network to solve the problem of knee injury during football training, which is beneficial to improve the training efficiency of football players [2]. Taking the soccer robot as the research object, the artificial neural network is used to construct the soccer robot behavior decision-making and basic action learning model, and the methods of improving artificial neural network is discussed to provide a certain foundation for football training. At the same time, in order to improve the selection of high-level football players, based on artificial neural network, the paper analyzes 23 comprehensive indicators of college football players, including the basic quality, body and function and so on, and offers the selection model of football players, which further lays a good foundation for college football training.