将粒子滤波算法引入到红外小目标的检测跟踪领域, 为解决重要性函数选择和粒子退化问题, 结合蚁群搜索算法与粒子群优化算法改进标准的粒子滤波。改进后的粒子滤波能有效减少粒子数, 降低运算量, 避免了粒子迁移线性化, 且无需重采样。分析论证了该算法的基本原理, 着重叙述了粒子迁移的过程。通过跟踪实拍红外图像以及仿真图像验证, 表明该算法能够成功跟踪信噪比低, 背景复杂, 运动非线性强的红外小目标, 对目标消失、遮挡情况也有一定的适应能力。

The difference of imaging characteristics between infrared polarization and infrared intensity are studied. A fusion algorithm of image is presented based on both support value transform (SVT) and local energy. The fusion result contains redundant information of images and increases the definition with more detailed information. Compared with the infrared polarization image and intensity image, the fusion image achieves the local standard deviation increase of 13.67% and 11.51%, local entropy increase of 16.46% and1.95% ,and average gradient increase of 15.41% and 44.05%, and the effectiveness of the algorithm is proved.