摘要
利用人工神经网络,无需对摄像机进行预标定,只要给出目标在计算机图像中的坐标点,就可直接得出摄像机云台水平和垂直需要调整的角度,将目标定位在图像中心。理论和实验证明,该方法比传统方法硬件要求低,算法简单,精度高。
Using artificial neural network, it can compute both the vertical and the horizontal rotation angles the pan-tilt-zoom camera needs to adjust when the object coordinate in the computer image is given. And what is more, it does not require the pre-calibration of the camera. The experimental error is about 10 pixels. Comparing with the traditional approaches, theoretical and experimental results show that this approach has more simple hardware and algorithm demands but more precision. It can be used in moving object tracking system.
出处
《计算机应用研究》
CSCD
北大核心
2006年第1期252-254,共3页
Application Research of Computers
基金
河北省博士基金资助项目(B2003202)
关键词
目标跟踪
定位
摄像机标定
人工神经网络
Object Tracking
Localization
Camera Calibration
Artificial Neural Network（ANN）
