Vector databases are revolutionizing the field of multi-camera tracking in video surveillance by enabling fast and accurate similarity searches across visual data. This approach addresses traditional challenges such as cross-camera identification, identity preservation, occlusion and disappearance, real-time processing requirements, privacy concerns, and scalability. Milvus, a purpose-built vector database, excels in this context with its ultra-fast similarity search capabilities, approximate nearest neighbor algorithms, scalable architecture, and flexible metric options. Its comprehensive feature set enables advanced tracking scenarios such as identity maintenance across challenging transitions, time-aware tracking, and identity resolution in crowded scenes. Real-world applications of Milvus include retail analytics, warehouse optimization, transportation hubs, and building your own tracking system. As vector database technology advances, we can expect even more sophisticated surveillance applications that combine diverse search capabilities for greater accuracy and insight.