基于分形和背景知识挖掘的红外图像水上桥梁目标识别  

Recognition of bridges over water in infrared image based on fractal theory and background knowledge mining

作　　者：闫晓珂[1] 史彩成[1] 赵保军[1] 何佩琨[1] 

机构地区：[1]北京理工大学信息科学技术学院电子工程系

出　　处：《系统工程与电子技术》2007年第7期1031-1033,1081,共4页Systems Engineering and Electronics

基　　金：国家部委预研课题(2002AA803032)

摘　　要：针对复杂背景下远距离航拍红外图像中水上桥梁识别的难题,提出一种新的基于分形理论和背景知识挖掘的目标识别方法。根据红外图像水上桥梁上下文描述,利用红外图像的直方图动态阈值法和分形特征检测出图像中的感兴趣区水域,同时基于先假设后检验的桥梁潜在目标被分割出来;然后利用形态学滤波完成目标图像的进一步分割;最后,利用目标特征匹配完成桥梁目标识别。基于TMS320C6416图像跟踪器的实验结果证明:该方法有较高的自动目标识别率和计算实时性。To recognize a large bridge over water under comptlcated background in long rang aerial rhotography infrared images, a novel three-step strategy based on fractal theory and background knowledge mining is presented. First, all regions in accord with the contextual description of the infrared images about bridges over water and regions of interest（ROD around the pixel 8-neighbors are detected based on fractal dimension and histogram active threshold, at the same time latent bridge targets are segmented. Then, image is performed over all sub images with morphological filters. Finally, the bridge target is recognized based on the characters matching. Experiments on the image tracker based on TMS320C6416 are conducted, and the results demonstrate that the approach performs well in automatic recognition accuracy and computational efficiency.

关 键 词：红外图像 桥梁 目标识别 数学形态学 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]