Review of Multi person Video Tracking Optimizations Using Optical Flow, Convolutional Neural Networks, and Marker less Motion Captures

The increasing need for accurate and efficient multi-person video tracking has led to much research in the optimization of tracking models for surveillance, healthcare, sports analytics, and behavioral analysis. However, despite the impressive progress made so far, the existing reviews within the domain rarely provide a structured taxonomy that can compare different models across key performance metrics such as accuracy, occlusion handling, real-time efficiency, and adaptability. Additionally, previous reviews lack a systematic analysis of deep learning-driven approaches, hybrid methodologies, and their integration with decentralized frameworks such as block-chain. This work fills these gaps by conducting a comprehensive, iterative taxonomy-based review of recent state-of-the-art multi-person tracking models, evaluating them under a PRISMA-driven framework process. This structured review provides an in-depth comparative analysis, identifying optimal models for real-time tracking, medical diagnostics, and blockchain-based decentralization. It has contributed to the development of next-generation tracking solutions by guiding researchers toward integrating hybrid AI models, decentralized computing, and energy-efficient tracking mechanisms that enhance tracking accuracy, security, and scalability in real-world scenarios.