基于概率支持向量机的激光雷达风切变图像的识别

The Recognition of Lidar Wind Shear Images Based on Probability Support Vector Machines

针对标准支持向量机在激光雷达风切变图像识别中无法提供后验概率这一问题,从有监督聚类的角度,提出一种基于FCM的概率支持向量机识别方法.先利用灰度-梯度共生矩阵提取激光雷达风切变图像的纹理特征,再利用支持向量机确定分类面,最后利用条件约束和FCM确定各类样本距离分类面的概率分布.实验结果表明,该算法对3种风切变的整体识别率可达到95.52%,与两种同类算法相比,识别率分别提高了1.27%和1.21%.

Aiming at the problem that the standard support vector machines （SVM） can not provide the posterior probability in the recognition of Lidar wind shear images, a modeling method of probability support vector machines based on fuzzy C-Means（FCM） was proposed. Firtly, gray-gradient co-occurrence matrix（GGCM） features was extracted from the laser radar images, then the classification face was determined by the support vector machines; at last, the constraints and FCM were adopted to determine the probability distribution of the samples. Experiment results show that the recognition rate of the proposed algorithm on the three kinds of low altitude wind shear can achieve to 95.52 %; compared with two kinds of similar algorithms, the recognition rate is raised by 1.27% and 1.21% respectively.