针对固定目标红外图像中边缘模糊,难以分割和识别的情况,提出了一种基于模板匹配的目标识别方法。由高程数据和正射影像等卫星数据生成目标区参考图和基准图,在基准图中根据归一化Laplace响应确定目标区特征尺度作为目标检测的先验知识,对实时图及灰度反转实时图进行匹配滤波,检测出候选区域,再对候选区域进行基于Hausdorff距离的模板匹配,从而得到最终识别结果。实验结果显示:该算法识别精度高、速度快,对于复杂地面目标前视红外图像的匹配识别具有一定的应用价值。

It is difficult to segmentation and recognition of FLIR image due to blurry edge, we introduced a recognition algorithm base on template matching. The reference image is created depending on the satellite dates and then created the template image, characteristic scale of target areas can be chosen as prior knowledge according to normalize LoG response. Using these scales, a number of blob-like candidate regions can be found with the same size with template image. Then the final result is obtained between candidate regions and the template image using Hasudorff measure. The experimental result shows that method have highly precision and quickly performance. It is useful to recognition of complex ground fixed garget.