| The low-level wind shear is a weather phenomenon which seriously threaten airliner flight safety. it’s charactered by happening suddenly, short duration, small-scale and strong strength, and have become a dangerous factor that seriously affect the plane’s takeoff and and landing. Because of its sudden happening and short duration, so the actual detecting materials are very lacking. Carrying on detection and recognition research onlow-level wind shear actively has important practical significance to ensure the safe flight of the airliner.First the status quo of wind shear and wind shear detection research are introduced, and then accomplishes the wind field modeling and simulation of different types of low-level wind shear with fluid dynamics software FLUENT, including the three-dimensional numerical simulation of micro downburst wind shear, the side wind shear and low-level jet.Then scanning the wind shear field by the analog laser radar in order to generate simulation radar scan data which will be converted to displayable radar images. The low-level wind shear sample library is build by scanning at different wind field position, in order to prepare for the identify of different types of wind shear.The feature extraction of wind shear images is an important step affect whether the recognition could be well proceeded. The wavelet transform has been widely used in image texture feature extraction, because of its characteristics of multi-resolution analysis and the ability to express characteristics of local signals in both the time domain and frequency domain. the wind shear regions of wind shear images are extracted by threshold segmentation;the step followed is two levels wavelet decomposition on it; the feature vectors are obtained by strike mean and standard deviation of each sub-band of wavelet coefficients.Finally, BP neural network is used to generate the classification results by identifying the eigenvector inputted. The simulation results show that the algorithm has a good feasibility.Finally, in feature recognition aspect, the texture feature vectors are identified and classified though BP neural network and support vector machines, the experimental results show that the method has good feasibility. |
