利用可见光图像进行军事侦察与制导时,目标往往掩盖在复杂背景中,难以准确识别,研究一种能准确提取目标特征的图像处理方法就成为图像精确制导的关键。文中提出了利用数学形态学的方法先确定目标范围,然后结合聚类方法进行小目标精确检测的算法。实验证明这种算法在背景复杂、目标可视面积较小的情况下,仍可以精确识别目标。该算法克服了聚类运算速度慢的缺点,从而能够快速的、准确地将目标提取出来,且具有良好的适应性。

In military reconnaissance and aerial guidance by visible light, targets often conceal in a complex background and difficult to identify, so it is of vital importance to exactly extract feature utilizing image processing methods. Based on the characteristics of background and target, it is put forward that morphology can be used to determine target scope, and further combined with the method of clustering to accurately detect small objects. The experimental results show that the method can realize accurate recognition even in complex background and relatively smaller visual target area. The method can overcome the low speed of clustering and quickly and accurately extract the target with good adaptability.