Lidar Imaging Recognition Of Low-level Wind Shear Based On Wavelet Invariant Moments

Low-level wind shear is an atmospheric phenomenon that has seriously influenced the aircraft flight, and it has different affect to flying along with the variation of types of wind shear as to the unique wind field characteristics. It takes enormously helpful to the pilot make the corresponding operation according to the correctly recognized wind shear.Therefore, it is actually significant and practically valued to research the classification of low-level wind shear actively. As the sudden appearance, short duration, small scale and related with the terrain and climate, the real low-level data is difficultly to detect and acquire. On the condition that ignoring the factors of specific terrain and climate, constructs the sample images of microburst low-level jet stream, side wind shear and tailwind-or-headwind shear through employing the computational fluid dynamic simulation technique combined the scanning mode of Doppler wind lidar, then recognize the types of low-level wind based on image features.Firstly, according to the properties of sample images in this four kinds of wind shear,chooses moment method which is not sensitive to the variation of shape of wind shear images to obtain the image features. Mainly studies the wavelet invariant moment method based on the Cubic B-spline, this feature extraction approach not only can effectively describe the global moment characteristics as radial velocity information of wind shear is relatively complete, but also local moment can get a precisely depict in order that merely use it to achieve a good recognition effect ultimately when there is a certain lack of radial velocity information. Then feature dimensions are reduced and optimized by Fisher Linear Discriminative Analysis with a low complexity so that the majorization features are fed into3-nearest neighbor classifier to identify four types of wind shear. There is a good recognition in the above means. Nevertheless, there is some space to further improve the recognition rate with regard to the selection way of wavelet moments. A new improved adaptive genetic algorithm(IAGA) is consequently proposed to pick up the best feature subset via analysing several previous IAGA at first. This new IAGA can control the evolution direction uniformly, and greatly maintain the population diversity simultaneouslyas to emphasize the fitness effect in individual and group. It is more suitable for the selection of wavelet moment critical features subset, which makes the wind shear recognition performance reach a steady and better result eventually.