一种基于时间Kalman估计器的红外小目标检测方法

An Algorithm Basedon Time-Domain Kalman Predictor for Infrared Small Targets Detection

目标在经过一个像素时,随着目标移动到该像素或离开该像素,该像素的灰度值会在出现一个冲击。据此提出了一种基于时间的 Kalman 单步估计器单像素处理算法,给出了 Kalman 估计器的递推方程。由前一时刻(前一帧)的估计结果和当前时刻的输入图像各点的灰度值,可以得到下一时刻(下一帧)图像各个像素点的背景估计,整个图像去除背景后就得到一幅包含目标的噪声图像,进而可以完成目标检测。采用红外图像序列进行了实验,实验结果验证了理论分析的正确性。最后得出了一些结论。

When a target traverses a pixel, an intensity pulse to the pixel will be bring. Based on this phenomenon, a time domain one-step single pixel Kalman predictor algorithm is put forward, and the recursive equations are also given. From the predicted result of a pixel of last moment and the input intensity of the pixel of this moment, the predicted value of the pixel intensity of next moment can be figured out. Removing the background(predicted value)of every pixel results an image consists of noise and targets, then targets detection can be done on this image. Experiments are carried out on infrared image sequence, and the results verify the analysis ahead. At last some conclusions are given.