基于最小一乘和遗传算法的红外弱小目标检测

在随机误差不服从正态分布的问题中,最小一乘估计的统计性能优于最小二乘估计;另外,最小一乘估计的稳健性更强。因此提出了基于最小一乘估计和遗传算法进行背景预测的红外弱小目标检测方法。首先,建立最小一乘准则背景预测模型,应用遗传算法求解最小一乘估计的最优值并进行背景预测;然后,由实际图像和预测图像相减得到残差图像,并采用二维指数熵图像阈值选取方法对残差图像进行分割。针对实际红外图像序列的实验结果表明:所提出的方法对弱小目标具有更高的检测概率和更好的检测结果,优于基于最小二乘背景预测的检测方法。

Infrared small target detection based on least absolute deviation and genetic algorithm

When the random error is not subject to normal distribution,the least absolute deviation estimation is superior to the least squares estimation.In addition,the robustness of the least absolute deviation estimation is also better than that of the least squares estimation.Thus,a method of weak and small target detection in infrared image sequences is proposed based on the least absolute deviation background prediction and the genetic algorithm.Firstly,a prediction model of the background signal based on the least absolute deviation criterion is founded.The extreme value is extracted by the genetic algorithm to predict the background.Then,the estimated image subtracted from the source image gives the residual image.The residual image is segmented using the threshold selection algorithm based on the two-dimensional exponent entropy.The experimental results with some real infrared image sequences show that the proposed method reduces the false alarm rate and greatly improves the detection performance of weak and small targets compared with the method based on the least squares background predication.