Target Detection and Tracking for Vehicles in Tourist Sites Based on Image Recognition

With the popularization of autonomous vehicles, reliable environment sensing and target recognition are the key to safe navigation of autonomous vehicles in locations with high human and vehicular traffic and many obstacles, such as tourist attractions. This study develops an optimized deep learning model for accurate detection and tracking of pedestrians and objects in complex tourist site environments based on image recognition. The model enhances the feature extraction capability of convolutional neural networks by incorporating an improved scene similarity perception module. A compact parallax map depth representation is utilized to identify targets and motion trajectories. The model is validated on a tourist site dataset and shows superior performance in dense crowd scenarios compared to existing methods. The reliable identification of pedestrians, vehicles and obstacles in various lighting and weather conditions ensures the safe navigation of autonomous vehicles and protects tourist safety.