Research and Implementation of Recognition Algorithm of Long-distance Runners Based on Deep Learning

The recognition of long-distance runners is mainly used to retrieve specific long-distance runner targets across video equipment in long-distance running events, which helps to improve the efficiency of the management of long-distance running events. At present, with the rapid development of artificial intelligence, lots of scholars utilize deep learning-based Person Re-Identification(Re-ID) technology to achieve long-distance runner recognition tasks. However, in practical applications, problems such as occlusion, noise, brightness changes, and color shifts usually affect the collected images of long-distance runners, thereby reducing the recognition accuracy of the existing Re-ID technology. For this reason, this paper proposes a recognition network for long-distance runners named Ldrr-net based on deep learning.Ldrr-net introduces the IGBN structure into the backbone network called Resnet50, which can reduce the adverse effects caused by the captured images, and has stronger robustness. In addition, we modify the loss, and propose Ldrr-loss to train network parameters, so that the network can better achieve intra-class aggregation and inter-class separation in the case of occlusion and similar features, and further improve the accuracy of long-distance runners' recognition. Experiments show that Ldrr-net has certain advantages in the recognition task of long-distance runners.