基于 SVM后验概率的红外弱小目标检测

摘要

To solve the problem of the disturbance of the background edge under complex cloud background,an infra-red small target detection method based on SVMposterior probability is proposed.In this method,infrared small target detection is considered as a binary classification problem between the target and the background.According to the characteristic of infrared image,the gradients in 8 directions of every pixel are taken as the classifying basis of the tar-get and background.And the gradients that can respectively reflect the characteristics of target and background are chosen to be the main part of the training set,so that the classification model can be acquired after training.The gradi-ents in 8 directions of every pixel are taken as the test set to obtain the posterior probability of SVM,which is exactly the output of detection,thus the targets are extracted from the image.The experimental results show that the algorithm is effective.%针对复杂云层背景中背景边缘干扰严重的问题，提出基于支持向量机（SVM）后验概率的红外弱小目标检测算法。该算法将红外弱小目标检测视作目标与背景的二分类问题，根据红外图像特性，以各像素点8个方向的梯度作为目标和背景的分类依据，选取能够表现目标和背景特征的梯度作为 SVM训练样本的主要参考量，设定训练集，并通过训练获得 SVM分类模型。基于 SVM后验概率的检测算法将待测样本各像素点的8方向梯度作用于分类模型，获得的 SVM后验概率作为检测输出。实验结果证明了该算法的有效性。