基于模板的低信噪比前视红外建筑物识别技术

针对低信噪比复杂地物背景下的前视红外地面建筑物目标自动识别问题，提出了一种基于目标模板的自动目标识别方法。首先，根据制备模板的尺度和形状信息对实时图像进行形态学背景抑制增强和目标重构处理；然后，对重构后的图像进行垂直和水平边缘线条提取和轮廓匹配，融合垂直和水平线条匹配结果得到最终的相关峰和潜在目标区域；接着，根据目标模板对潜在目标区域进行区域灰度对比度度量；最后，融合轮廓匹配和区域灰度对比度度量结果，得到最终相关峰，实现目标的识别。大量数据测试表明，该方法的正确识别率在92%以上，定位误差小于2个像素，速度快、检测性能好、适应性较强、易于硬件实现。 

Method of target recognition from low SNR building FLIR images based on template

In order to solve the problem of automatic recognition of low SNR FLIR ground building in complex background, the paper proposes a target recognition approach based on target template. First, according to the prepared template, the real-time images are processed by morphology background suppression and target reconstruction. Second, by processing the reconstructed images with edge extraction and contour matching, the matched correlation peak matrix and ROI are attained. Then, RGC measurement is calculated in the ROI region according to target template. Finally, by integrating contour matching and RGC measurement results, the ultimate correlation peak matrix is obtained and the target recognition is achieved. Large amounts of data tests manifest that the rate of accurate recognition is above 92%, the location error is less than 2 pixels. The test results show this approach has advantages in speed, detecting performance and adaptability, which is easy to implement in hardware as well.