The Study On Identification Of Wind Shear Based On Image Multiscale Analysis

Low-level wind shear seriously affects the efficiency of ATC, threatening the safety of air traffic. Different types of wind shear influence on aircraft in different way, to return to planted route, pilots must take corresponding measures against wind shear type. Effective type identification of low-level wind shear is of great significance for ensuring flight transportation security.Base on research on low-level wind shear, four typical wind farm models of crosswind, headwind, low level jet and downdraft were selected in this paper, then the type identification algorithm from the perspective of image multi-scale analysis was proposed. Details are as follow:Firstly, the four general ideal models of low-level wind shear farm were constructed by FLUENT software engineering modeling method and detected to get PPI images of the low-level wind shear based on Lidar VAD working mode, then the Lidar low-level wind shear sample library was built. The library contains sample images of wind radial velocity information not only complete but also partly missing to provide support for further research on type recognition of low-level wind shear.For the anisotropy of wind shear image samples, an identification algorithm of low-level wind shear was proposed based on image multi-scale analysis. Count the energy and standard deviation on every scale of L-NSCT images as feature vector to get rotation invariant texture feature. The proposed method overcomes the sensitive to image rotation transform in traditional NSCT, and the extracted features can reflect the frequency distribution of different types of wind shear. After training by support vector machine, the results show the L-NSCT texture feature extraction algorithm can improve the low-level wind shear type recognition rate compared with some of the current low-level wind identification method.Considering both spatial feature and frequency feature can perform the type characteristics of low-level wind shear, a novel identification method was proposed based on feature fusion through combining the L-NSCT approach above and WLD histogram feature extracted from the spatial information of wind shear farm. Then put the final wind shear texture feature vectors into SVM, the recognition rate is significantly higher than the existing low-level wind shear recognition method based on shape feature.