In this paper, residual neural network is adopted to extract Chu culture elements, and in order to reduce the occurrence of overfitting phenomenon, ResNet is improved based on random deactivation strategy. After that, the characteristics of Chu culture elements are integrated with sports digitization. The data source, experimental environment, and evaluation index are determined to test the extraction and integration effect. It is found that the ResNet network model in the sports culture video of Chu culture elements image classification accuracy reached 91.43%, based on the results of the analysis of the difference between the integration effect of Chu culture elements before and after the intervention, verifying the integration effect of Chu culture elements This paper not only allows the Chu culture elements to be inherited, but also makes the sports culture digitalization level of development to be further improved.