Optimizing ROI Benefits Vehicle ReID in ITS

Vehicle re-identification (ReID) matches the same vehicle across different cameras in a surveillance system, crucial for Intelligent Transportation Systems (ITS). This study investigates if optimal detection regions, guided by confidence scores, enhance feature matching and ReID tasks. Using YOLOv8 for detection and DeepSORT for tracking across twelve Indiana Highway videos, vehicle images were cropped from inside and outside Regions of Interest (ROIs). Features were extracted using ResNet50, ResNeXt50, Vision Transformer, and Swin-Transformer. Results showed higher cosine similarity for in-ROI images, and the most significant difference was observed during night conditions (0.7842 inside vs. 0.5 outside the ROI with Swin-Transformer) and in cross-camera scenarios (0.75 inside-inside vs. 0.52 inside-outside the ROI with Vision Transformer). Information entropy and clustering variance supported greater feature consistency in ROIs. These findings suggest strategically selected ROIs can improve tracking and ReID accuracy in ITS.