基于邻域对比的红外林火小目标检测

Small Infrared Forest Fire Target Detection Based on Neighborhood Contrast

为了在森林复杂背景下准确地检测出红外小目标,提出了一种基于邻域对比的目标提取算法。首先,根据小目标区域与其8-邻域背景的差异,利用邻域对比算法实现对小目标的增强和对背景的抑制;其次,采用多尺度模板准确检测小目标区域的变化情况;最后,在得到最终对比图的基础上,利用自适应阈值对目标进行分割。实验结果表明:与现有算法相比,所提出的算法在红外小目标检测方面具有更高的准确性,图像整体的信噪比也有较大的提高。

In order to accurately detect the small infrared target at the forest complex background, an algorithm for target extraction was proposed based on neighborhood contrast. First of all, according to the difference between small target area and its 8-neighborhood background, the small target enhancement and the background suppression were obtained by using of the neighborhood contrast algorithm. Secondly, a multi-scale template was employed to accurately detect the small target in the gray varied area. Finally, based on final comparison result, we adopt an adaptive threshold for the target segment. Experiments proved that the proposed algorithm has more accuracy than the existing algorithms in small infrared target detection, and the whole image SNR has been greatly improved simultaneously.