Abstract
Accurate detection and tracking of small objects such as pedestrians,cyclists, and motorbikes are critical for traffic surveillance systems, whichare crucial in improving road safety and decision-making in intelligenttransportation systems. However, traditional methods struggle with challengessuch as occlusion, low resolution, and dynamic traffic conditions,necessitating innovative approaches to address these limitations. This paperintroduces DGNN-YOLO, a novel framework integrating dynamic graph neuralnetworks (DGNN) with YOLO11 to enhance small object detection and tracking intraffic surveillance systems. The framework leverages YOLO11's advanced spatialfeature extraction capabilities for precise object detection and incorporatesDGNN to model spatial-temporal relationships for robust real-time trackingdynamically. By constructing and updating graph structures, DGNN-YOLOeffectively represents objects as nodes and their interactions as edges,ensuring adaptive and accurate tracking in complex and dynamic environments.Extensive experiments demonstrate that DGNN-YOLO consistently outperformsstate-of-the-art methods in detecting and tracking small objects under diversetraffic conditions, achieving the highest precision (0.8382), recall (0.6875),and [email protected]:0.95 (0.6476), showcasing its robustness and scalability,particularly in challenging scenarios involving small and occluded objects.This work provides a scalable, real-time traffic surveillance and analysissolution, significantly contributing to intelligent transportation systems.