由于多种因素的影响，红外小目标容易淹没在背景中，日趋复杂的背景以及隐身技术的应用，给红外小目标的检测带来极大的困难。分析天空背景下的红外图像特征，针对红外目标、天空背景和噪声各自不同的特点，结合二维信息熵，基于聚类思想构造属性集，依据属性集进行目标检测，然后根据噪声和目标点的特征，对检测结果中的孤立噪声点进行剔除，最终检测出小目标。仿真实验结果表明：该算法较二维最大熵算法，能够有效地提取出目标，与二维灰度级-邻域灰度级绝对差直方图法相比，计算简单方便，缩短运行时间。

Due to many factors, small infrared targets easily drown in the background. Additionally, increasingly complex backgrounds and stealth technology applications create abundant difficulties for infrared target detection. Infrared images are analyzed from the characteristics of sky background, noise and target, and two-dimensional information entropy. This is combined with a clustering structure attribute set, and the isolated noise points are then removed, pernoise and target point characteristics of the test results. Finally, the small target is detected. Simulation results show that, compared with the two-dimensional Maximum Entropy algorithm, the proposed algorithm can effectively find the target. Compared with the two-dimensional gray-scale neighborhood gray-level histogram method, calculation is simple and convenient, and requires a short running time.