Multi-camera Environment, Target Handover Research And System Realization

Looking and analyzing human is a research focus in computer vision for its wide applications, such as intelligent vision surveillance, human behavior understanding, gait recognition and virtual reality and so on. However the limitation of camera's resolution and Field Of View (FOV) constrains the development of monocular camera based human motion analysis. It makes the use of multiple cameras system (MCS) to be necessary. There are several key techniques such as continuous target tracking and data communication between multiple cameras, collaborative work of multiple cameras and recovering FOV lines of multiple cameras being studied in this paper. The main works are described as following:1. Analyzing and studying the key techniques and correlative algorithms in multiple cameras human tracking system in detail. Summarizing the primary technical difficulty and algorithms studied currently, these provide valuable reference for future works.2. A new algorithm for generating camera FOV lines fast and automatically, synchronized video based, is presented in this paper. To generate the camera's FOV lines fast, exactly, real-time and automatically the algorithm just needs a piece of synchronized video without any camera's calibration and environment information. This provides an effective tool for future work like field of view segmenting and target handoff between multiple cameras.3. An improvement of the computation of the distance between camera FOV line and target is adopted to improve the accuracy of target handoff and the time of computation. Experiments indicate our algorithm was faster and easer for resolving target handoff problem with high veracity in multiple cameras human tracking field.4. Developing a system of multiple cameras human detection and tracking based on PC. A data transmission technique based on Socket is adopted in computer communication to adapt to transmit message real-time, and before data transmission a data coding work was done to improve data transmission security and data varieties. The system has several functions as follows: target handoff between multiple cameras, generating camera FOV lines automatically, active camera control, multiple clients communicates with each other real-time. We also separately design thehuman-machine interactive interface for client and server, the interface is very friendly and convenient for users.