基于框架的远距红外桥梁目标识别  

A Recognition Algorithm Based on Knowledge Framework for Long range Infrared Bridge Images

作　　者：孙琪[1] 曹治国[1] 张天序[1] 

机构地区：[1]华中科技大学图像识别与人工智能研究所,图像信息处理与智能控制教育部重点实验室

出　　处：《华中科技大学学报（自然科学版）》2001年第4期1-3,共3页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目!(6 9875 0 0 5 )

摘　　要：针对远距红外图像中桥梁目标的特征 ,提出了一种基于知识框架的桥梁目标识别算法 .由于远距红外桥梁图像背景复杂、干扰大 ,同时目标在图像中所占区域小 ,采用基于整幅图像的全局算法是不现实的 .因此 ,依据桥梁的特征 ,先检测潜在的桥梁点区域 ,然后对这些局部区域进行考察分析 ,根据其与预先建立的知识框架实例匹配程度的不同赋相应的置信度 ,从而去除虚假目标 ,找到真正的桥梁 .Through the analysis of the features of bridge object in long range IR image, a new recognition algorithm based on knowledge framework is put forward. Due to some characters of long range infrared bridge images, the global method processing the whole image equally is ineffective. Based on the features of the bridge object, our proposed algorithm first detects the candidate bridge regions and then focuses on them. According to the degree to which they match our prebuilding framework, different credits are given, so the false targets are excluded and eventually the real target is found. The experiment results demonstrate the good performance of our method.

关 键 词：目标识别 红外图像 知识框架 置信度 桥梁检测 

分 类 号：U446[建筑科学—桥梁与隧道工程]