CAJ | 학술논문

针对微下击暴流、低空急流、顺逆风以及侧风低空风切变样本图像间的形状特性关系，主要研究了小波不变矩的特征提取技术在风切变识别中的应用。首先，采用基于三次B样条的小波不变矩提取风切变图像的形状特征。然后，将提取的特征通过Fisher线性判别分析(LDA)降低维数，实现风切变有效特征的提取。最后，采用三阶近邻分类器分类识别四种低空风切变。实验结果表明，该算法与应用Hu矩和Zernike矩特征进行分类识别相比，识别结果更加稳定，且平均识别率得到了较大提高，能够有效用于风切变图像的类型识别中。
침대미하격폭류、저공급류、순역풍이급측풍저공풍절변양본도상간적형상특성관계，주요연구료소파불변구적특정제취기술재풍절변식별중적응용。수선，채용기우삼차B양조적소파불변구제취풍절변도상적형상특정。연후，장제취적특정통과Fisher선성판별분석(LDA)강저유수，실현풍절변유효특정적제취。최후，채용삼계근린분류기분류식별사충저공풍절변。실험결과표명，해산법여응용Hu구화Zernike구특정진행분류식별상비，식별결과경가은정，차평균식별솔득도료교대제고，능구유효용우풍절변도상적류형식별중。
According to the shape characteristic relationship within microburst, low﹣level jet stream, side wind shear and tailwind﹣or﹣headwind shear images, the feature extraction technique of wavelet invariant moment applied to the recognition of wind shear was mainly studied. Firstly, wavelet invariant moments method was employed to extract shape features of low﹣level wind shear images, which was based on cubic B- spline wavelet basis. Then, the feature dimensions were reduced by Fisher Linear Discriminative Analysis (LDA) in order to get the effective shape features of target images. Finally, the effective shape features were fed into 3- nearest neighbor classifier to identify four types of low﹣level wind shear. The experiment results demonstrate that the proposed approach has stronger robustness and better average recognition rate compared to the recognition effect based on Hu moment and Zernike moment, which can effectively be used to recognize the type of wind shear images.