The Study On Identification Of Wind Shear Based On Laser Radar

Low-level windshear has become a serious factor in the process of airplane taking off and landing, at the same time, different type of windshear has its unique feature, so they have different influence on the flight of airplane, so it needs the pilot makes correct operation to different type of windshear. So the detection of Low-level windshear and the identification of its type will be very important.In the paper, The Lidar images of windshear are simulated, and the research of identify four types of windshear has been done, such as micodownburst, low-level jet, sea breeze,crosswind. Details are as follows:First, For the lack of Lidar data of windshear, Windshear are simulated based on Fluent numerical simulation method, in the way, The basic characteristics of four types of windshear under ideal conditions are gotten; and referencing to Hong Kong airport Lidar detection mode,while directly rotating the wind farm, finally the Lidar images of windshear under different wind directions are gotten.Second, For the variations of Lidar radial projection of wind speed under different wind directions and each type of windshear has its inherent shear strength in different regions, A identification method of a combination of LBP local texture features and Gray-Gradient Co-occurrence Matrix overall texture features is proposed. LBP local texture features is not sensitive to the changes of variations of wind speed projection, Gray-Gradient Co-occurrence Matrix overall texture features stand for the shear strength relationship of the whole windshear region, and then through CCA, the combination features of the two texture are gotten; finally the combination features of the four types of windshear are identified through SVM method. Final experiment show that the proposed method get a better performance than the single texture features.Third, For the problem that the standard SVM does not provide the posterior probability in the recognition of Lidar wind shear images, a modeling method of probability support vector machines based on Fuzzy C-Means(FCM) is proposed, compared with the traditional method, there are some performance improvement.