摘要
为了在实时视频监控中能够有效地检测和跟踪目标,提出了一种基于区域的图像拼接算法。该算法根据空间几何知识建立摄像机旋转投影模型,对需要拼接的图像进行二维空间坐标变换,使所有的图像都转换到同一坐标系下。最后,根据已知的云台旋转参数来设定配准特征空间,进而对图像进行配准和拼接。算法中建立了一种新的摄像机旋转投影模型,可以有效地进行图像的空间坐标转换,并且通过设定配准特征空间,有效地改进了配准和拼接算法。在PC机上的实验结果表明了该算法的准确性和有效性。
In order to real-time video surveillance can effectively detect and track targets, a region-based image stitching algorithm is proposed for the realization for monitoring using of PZT （photographic zenith telescope） and cameras. First, the camera rotating projection model is proposed based on the spatial knowledge of geometry; Based on this model, the images need to transform in the two-dimensional space coordinates, so that all images are converted into the same coordinate system. According to the known rotation PTZ＇ s parameters to set the registration feature space, and then registering and stitching the image. Algorithm presents a new projection model of camera rotation to effectively solve the spatial coordinate transformation, and according to set the registration feature space, the image registering and stitching algorithm is greatly improved. Finally, by the experiment on the PC, the validity and efficiency of the presented method is demonstrated.
出处
《计算机工程与设计》
CSCD
北大核心
2010年第18期4044-4046,4050,共4页
Computer Engineering and Design
基金
上海市重点学科建设基金项目(J50103)
上海市教育委员会科研创新基金项目(08YZ16)
上海大学创新基金项目(A.10-0108-07-002)
关键词
运动检测
摄像机投影
图像拼接
图像配准
特征空间
moving detection
camera projection
image stitch
image registration
feature space
