State-of-the-art and research gap analysis of artificial intelligence based techniques for automatic vehicle recognition

Automatic Vehicle Recognition (AVR) has become a cornerstone technology for Intelligent Transportation Systems, smart city infrastructure, traffic monitoring, and border security. Recent advancements in Artificial Intelligence (AI), particularly deep learning, have markedly improved vehicle detection, classification, and tracking. However, real-world deployment of AVR systems remains challenged by factors such as high computational demands, limited cross-domain generalization, scalability constraints, and the need for privacy-preserving operation. This paper presents a comprehensive state-of-the-art review and critical gap analysis of AI-based vehicle recognition techniques. We systematically examine contemporary approaches, including CNN-based one-stage and two-stage object detectors, transformer-based architectures with global self-attention, fine-grained vehicle classification models, and vehicle re-identification methods. Additionally, we explore multimodal sensor fusion strategies and Edge AI deployment to evaluate their effectiveness in enhancing robustness and real-time performance under diverse environmental conditions. The analysis identifies key limitations in current systems, such as the absence of unified end-to-end frameworks, the high computational cost of transformer models for edge deployment, insufficient generalization across geographic regions and vehicle types, and limited explainability and privacy safeguards. Finally, we outline promising research directions aimed at developing lightweight, adaptive, and privacy-conscious AVR systems capable of bridging the gap between laboratory research and large-scale real-world applications.