| Summary: | 碩士 === 國立臺灣科技大學 === 自動化及控制研究所 === 100 === For traditional monitoring system, it still using passive camera architecture, its viewing angle is fixed and the scope is also limited. No matter the invader appeared or not, it can only provide the recording function. This study proposes a real-time invader tracking and location system by cross-eye architecture. The active camera can rotation in pan and tilt direction, reaching a wide range of monitoring, through the computer vision to analyze the captured image, so that the system can initiatively detect and track invader This study can divide into the detection procedure, tracking procedure and locating procedure. Firstly, in the detection procedure, the two cameras rotating in pan direction continuously, used to detect the invader. Until the invader appeared in which a camera screen, the system would calculate the centroid of the invader, then sending the commands to control the optical axis of the two cameras to aim the invader. For tracking procedure, the two cameras were rotating different angles to track the invader, ensure the invader locate at the center of the image. Finally, the system computed two cameras' rotation angles in locating procedure, using the parallax principle to calculate the depth information of the invader. By the depth information, the system calculate the three-dimensional coordinates of the invader.
The experimental result shows that, this study can effectively detect the invader and track it in a complex background. In addition, different to other studies of stereo vision, the middle of two cameras is blind area, this system is not limited to that, but can be effectively calculate three-dimensional coordinate of the invader.
|
|---|
