An Improved CrowdDet Algorithm for Traffic Congestion Detection in Expressway Scenarios

Traffic congestion detection based on vehicle and tracking algorithms is one of the key technologies for intelligent transportation systems. However, in expressway surveillance scenarios, small size occlusion present severe challenges this method, including low accuracy traffic accuracy. To address these challenges, paper proposes an improved version CrowdDet algorithm by introducing Involution operator bi-directional feature pyramid network (BiFPN) module, which called IBCDet. The proposed IBCDet module can achieve higher scenarios enabling long-distance information interaction multi-scale fusion. Additionally, a vehicle-tracking designed to calculate running speed vehicles, it uses average according Chinese level serviceability (LoS) criteria. Adequate experiments are conducted both self-built Nanjing Raoyue monitoring video dataset (NJRY) public UA-DETRAC. experimental results demonstrate that outperforms commonly used object