Answers are grounded in this paper and cite their sources.
The rapid increase in vehicle numbers has exacerbated challenges in modern transportation, leading to traffic congestion, accidents, and operational inefficiencies. Intelligent Transportation Systems (ITS) leverage computer vision techniques for vehicle detection, improving safety and efficiency. This paper aims to provide a comprehensive review of vehicle detection methods in ITS. Traditional image-processing techniques, including Scale-Invariant Feature Transform (SIFT), Viola-Jones (VJ), and Histogram of Oriented Gradients (HOG), are analyzed. Additionally, modern deep learning-based approaches are examined, distinguishing between two-stage methods such as R-CNN and Fast R-CNN, and one-stage methods like YOLO and SSD. Various image acquisition techniques, including Mono-vision, Stereo-vision, Thermal/Infrared Cameras, and Bird’s Eye View, are also reviewed. The analysis highlights the evolution from handcrafted feature-based methods to deep learning techniques, demonstrating significant improvements in detection accuracy and efficiency. One-stage detectors, particularly YOLO and SSD, offer real-time performance, while two-stage methods provide higher precision. The impact of different imaging modalities on detection reliability is also discussed. Advances in deep learning and imaging techniques have significantly enhanced vehicle detection capabilities in ITS. Future research should focus on improving robustness in diverse environmental conditions and optimizing computational efficiency for real-time deployment.
No takes yet. Share an insight, caveat, or question.
Asri et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: