激光雷达中基于组合纹理的低空风切变识别算法

发布时间：2019-08-06 12:20

【摘要】：针对不同的风切变在激光雷达图像上所呈现的不同纹理特性,提出了一种组合局部纹理特征和全局纹理特征的识别方法。先分别从激光雷达风切变图像中提取LBP特征和灰度-梯度共生矩阵特征,LBP特征反应图像的局部纹理,代表风场局部风速的变化,灰度-梯度共生矩阵特征反应图像的全局纹理,代表风场全局的风速变化,再通过典型相关分析对两种特征进行融合,最后采用最近邻分类器对三种风切变进行匹配识别。实验结果表明,该算法对三种低空风切变的平均识别率达到99.02%,与三种单一的纹理特征分类识别相比,分别提高了18.86%,5.88%和7.01%。
[Abstract]:Aiming at the different texture characteristics of different wind shear on lidar images, a recognition method combining local texture features with global texture features is proposed. Firstly, LBP features and gray-gradient symbiosis matrix features are extracted from lidar wind shear images, LBP features reflect the local texture of the image, represent the local wind speed change of the wind field, the gray-gradient symbiosis matrix feature reflects the global texture of the image, and represent the global wind speed change of the wind field. Then the two features are combined by canonical correlation analysis. Finally, the nearest neighbor classifiers are used to match and recognize the three kinds of wind shear. The experimental results show that the average recognition rate of the algorithm for three kinds of low altitude wind shear is 99.02%, which is 18.86%, 5.88% and 7.01% higher than that of the three single texture feature classification.
【作者单位】：民航气象研究所中国民航大学;智能信号与图像处理天津市重点实验室中国民航大学;
【基金】：国家自然科学基金(No.41075013) 中央高校基金(No.ZXH2010D020,3122013P009) 中国民航大学研究生科技创新基金(No.3122013SY25)资助
【分类号】：TN958.98

【参考文献】