基于小波矩的激光雷达成像低空风切变识别

发布时间：2018-05-22 12:51

本文选题：风切变 + 计算流体力学　；参考：《中国民航大学》2014年硕士论文

【摘要】：低空风切变是一种严重影响飞机飞行的大气现象,不同风切变类型有其各自不同的风场特征,对飞行的影响也大有不同。正确识别风切变类型,能为飞行员做出相应的操作带来极大的帮助。因此,积极开展对低空风切变的识别具有重要的现实意义和实用价值。低空风气变具有突发性、持续时间短、尺度小等特点,并与实际的地形、气候等相关,使其实际数据探测比较困难。在忽略地形、气象因素条件下,采用计算流体力学(Computational Fluid Dynamic,CFD)仿真软件结合多普勒测风激光雷达的扫描方式,构造了微下击暴流、低空急流、顺逆风以及侧风低空风切变的样本,基于图像特征对低空风切变进行分类识别。首先,根据微下击暴流、低空急流、顺逆风以及侧风低空风切变的样本特性,选择对图像形状关系变化不敏感的矩方法提取风切变特征。主要研究了基于三次B样条小波基的小波矩形状特征提取方式,使得不仅在图像径向速度信息较全面时,较好地描述风切变的全局矩特征;同时,在图像样本径向速度信息存在一定缺失度时,也能精确地刻画其未缺失的局部特征,达到较好地识别性能。为验证小波矩特征提取算法的有效性,先采用复杂度较低地Fisher线性判别方式(Linear Discriminative Analysis,LDA),最大化样本可分性降低小波矩维数,送入三阶近邻识别四种低空风切变。其次,在此基础上进一步研究了小波矩特征的选择,目的是再次提高风切变的识别性能。通过分析对比几种已有的改进自适应遗传算法(Improved Adaptive Genetic Algorithm,IAGA)的优缺点,创建了一种新的改进自适应遗传算法。该算法在均匀把握种群进化方向时,根据个体在当代群体中的作用丰富种群的多样性,更适于选择小波矩的最优特征子集,使风切变达到了一个稳定、较优地识别效果。
[Abstract]:Low level wind shear is an atmospheric phenomenon which seriously affects the flight of aircraft. Different types of wind shear have their own characteristics of wind field and have different effects on flight. The correct identification of wind shear type can greatly help the pilot to make the corresponding operation. Therefore, it has important practical significance and practical value to actively carry out the identification of low-altitude wind shear. The low-altitude gas change is characterized by sudden occurrence, short duration, small scale and so on, which is related to the actual terrain and climate, which makes it difficult to detect the actual data. Under the condition of neglecting the terrain and meteorological factors, using computational fluid dynamics Fluid dynamic CFDs and the scanning mode of Doppler wind lidar, the samples of micro-downburst flow, low-altitude jet flow, headwind and crosswind low-altitude wind shear are constructed. Classification and recognition of low altitude wind shear based on image features. Firstly, according to the sample characteristics of micro-downburst, low-altitude jet, headwind and crosswind low-altitude wind shear, the moment method which is not sensitive to the change of image shape relationship is selected to extract the wind shear feature. Based on cubic B-spline wavelet basis, wavelet rectangular feature extraction method is studied in this paper, which not only describes the global moment feature of wind shear better when the radial velocity information of image is more comprehensive, but also, When the radial velocity information of the image sample has a certain degree of deficiency, it can also accurately describe the local features that are not missing, and achieve better recognition performance. In order to verify the validity of the wavelet moment feature extraction algorithm, linear Discriminative analysis is used to maximize the sample separability and the wavelet moment dimension is reduced by using the lower ground Fisher linear discriminant method. The wavelet moment dimension is reduced and the third order nearest neighbor is sent to identify four kinds of low-altitude wind shear. Secondly, the selection of wavelet moment features is further studied in order to improve the performance of wind shear recognition again. By analyzing and comparing the advantages and disadvantages of several existing improved Adaptive Genetic algorithms, a new improved adaptive genetic algorithm is proposed. This algorithm is more suitable for selecting the optimal feature subset of wavelet moments according to the function of individual in the contemporary population, which makes the wind shear reach a stable and better recognition effect.
【学位授予单位】：中国民航大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.52

【参考文献】