基于形态学梯度的红外目标检测

对于红外跟踪系统中的低空非合作运动目标检测问题,由于距离未知,目标可能表现为小目标或者面目标,另外,红外图像中可能包含大量的地物背景.针对以上问题,本文提出了一种基于形态学梯度的目标检测算法.此方法首先计算红外图像的形态学梯度,进而根据各种地物背景的形态学梯度特征提取天空区域,然后在此区域内进行目标检测,最后进行目标边缘片段的合并.该方法基于形态学梯度对于灰度变化的敏感性,使用形态学梯度来表征红外目标边缘强度,通过检测目标边缘来达到检测目标的目的.实验证明,本文方法能够有效的检测目标,虚警较少,且对小目标和面目标具有一定通用性.

IR Target Detection Based on Morphological Gradient

We address the problem of detecting targets with different areas in air background, which will be very complex when containing part of ground. A target detection approach based on morphological gradient is proposed. Firstly, morphological gradient of the IR frame is computed. Then the air region is gotten based on the morphological gradient. At last, targets in the air region are detected and labeled. Because morphological gradient is sensitive to the slow changing grayscale on IR targets’ boundaries, morphological gradient is used to strengthen targets’ boundaries and detect targets by detecting the boundaries. Our experiments show that the method yields good detecting performance and is suitable for detecting targets with different areas.