基于图像多尺度分析的低空风切变识别研究

发布时间：2018-05-26 20:45

本文选题：低空风切变 + 图像处理　；参考：《中国民航大学》2015年硕士论文

【摘要】：机场周围低空风切变严重影响空中交通效率,威胁空中交通的安全。低空风切变的风场具有多种表现形态,其对飞行的影响相差甚远,驾驶员需要参考其类型采取与之对应的操作。研究低空风切变类型判别课题对于航空航天来说是十分必要的。本文在国内外对低空风切变研究的基础上,选取低空侧风、低空逆风、低空急流和下冲气流四种典型的低空风切变风场进行建模,从图像多尺度分析的角度研究低空风切变的类型识别方法。具体工作如下:首先,文中通过FLUENT的工程化建模方法构建上述低空风切变风场的一般理想模型,采用基于激光雷达速度方位工作模式扫描风场模型得到低空风切变的PPI图像,构成激光雷达低空风切变样本库,库中含有风场径向风速信息完整和缺失的样本图像,为进一步研究低空风切变类型识别提供支持。其次,针对风场样本图像切变线的各向异性,本文选择了一种基于图像多尺度分析理论的低空风切变类型识别方法。统计每个L-NSCT变换尺度下的低空风切变各子代的能量和标准差作为特征向量,得到对风场旋转不变的纹理特征。文中方法克服了传统多尺度分析方法对图像平移旋转变化的敏感,提取到的特征能够反映不同类型低空风切变风速的频率分布。经过支持向量机训练识别的结果表明,与目前的一些低空风切变类型识别方法相比,本文采用的L-NSCT旋转不变纹理特征提取算法提高了低空风切变类型识别率。最后,考虑到频率特征和空域特征均能表现低空风切变的类型信息,在上文L-NSCT方法的基础上,利用WLD获取低空风切变风场的空域信息,选择了一种基于特征融合的识别方法,获得最终的低空风切变纹理特征向量,送入SVM中训练识别,识别性能明显高于现有的基于形状特征的低空风切变识别方法。
[Abstract]:The low-altitude wind shear around the airport seriously affects the air traffic efficiency and threatens the safety of air traffic. The wind field of low altitude wind shear has many forms and its influence on flight is very different. The pilot needs to refer to the type of wind shear and take the corresponding operation. It is necessary for aerospace to study the problem of low-altitude wind shear type discrimination. On the basis of the research on low level wind shear at home and abroad, four typical low level wind shear fields, namely, low level crosswind, low level headwind, low level jet flow and downwind flow, are selected to model the wind field in this paper. In this paper, the classification recognition method of low level wind shear is studied from the view of image multi-scale analysis. The specific work is as follows: firstly, the general ideal model of the wind shear wind field is constructed by the engineering modeling method of FLUENT, and the PPI image of the low altitude wind shear is obtained by scanning the wind field model based on the velocity azimuth mode of the lidar. A low altitude wind shear sample library of lidar is constructed, which contains complete and missing sample images of wind field radial wind speed information, which provides support for further research on low altitude wind shear type recognition. Secondly, in view of the anisotropy of wind field sample image shear line, a low level wind shear type recognition method based on image multi-scale analysis theory is selected in this paper. The energy and standard deviation of each generation of low-altitude wind shear at the scale of L-NSCT transform are calculated as feature vectors, and the texture features with invariable rotation to wind field are obtained. The method overcomes the sensitivity of traditional multi-scale analysis method to image translation and rotation, and the extracted features can reflect the frequency distribution of different types of low-altitude wind shear wind speed. The results of SVM training and recognition show that the L-NSCT rotation invariant texture feature extraction algorithm proposed in this paper improves the recognition rate of low-altitude wind shear type compared with some current low-altitude wind shear type recognition methods. Finally, considering that the frequency feature and the spatial feature can represent the type information of the low-altitude wind shear, on the basis of the L-NSCT method above, using WLD to obtain the spatial information of the low-altitude wind shear field, a recognition method based on feature fusion is selected. Finally, the final low-altitude wind shear texture feature vector is obtained, and then trained in SVM. The recognition performance is obviously higher than the existing low-altitude wind shear recognition method based on shape features.
【学位授予单位】：中国民航大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：V321.225;TP391.41

【相似文献】