Intelligent objects tracking with automatic image zoom in/out for real-time video surveillance systems

• Main Authors: Jhong-mao Wei, 魏仲懋
• Other Authors: Shih-chang Hsia
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/62667275877404360309

碩士 === 國立雲林科技大學 === 電子與光電工程研究所碩士班 === 101 === Besides failing to take panorama photo-shoots due to the fixed angle of cameras, general traditional cameras are also unable to track and detect human faces. On the contrary, the application of active cameras can create a better monitoring system through an expanded range of monitor and an improved resolution.This paper puts forward an algorithm that can automatically detect, track, zoom in, and zoom out objects, as well as use active camera to transmit images from an image capture card to a computer to conduct image processing. In motion detection, the temporal differencing method is used to identify the location of a moving object, and further confirm the coordinate points of the tracked points on the face by detecting the hair and skin colour of the located object and control the rotating angel of the active camera through RS232 serial communication, obtain the tracking result by adjusting the rotating speed of the camera in accordance with the speed of the moving object, and finally use the zooming-in and zooming-out function of the active camera to control zooming-in of the more distant objects. The proposed detection and tracking algorithm in this paper can keep the person being tracked at the central range of the screen When the size of the moving object is smaller than the threshold set by us, the system will automatically adjust the magnification of the active camera to 2x, 4x, 6x, and 8 x zooms respectively according to the threshold, to capture a clearer photo of human face.