基于Fisher两类判别的机库目标识别

为实现根据波谱特征识别高分可见光遥感影像上的机库目标,通过分析影像上机库与其他常规地物之间波谱特征的区别,利用常规的Fisher两类判别法分类识别,并针对错分像元较多的问题构建逐层剥离法进行改进。结果表明:逐层剥离法可识别出77%以上的机库目标像元,错分像元较常规方法减少85%以上,能有效去除影像上大部分其他地物的干扰,具有更高的识别精度和效率。

Hangar Target Recognition Based on Fisher's Two Kinds of Discrimination

In order to recognize the hangar target in high-resolution visible remote sensing image based on spectral features, the difference of spectral features between hangar and other conventional objects was analyzed, and the conventional Fisher discriminant method was used to classify and identify the hangar target, and a layer-by-layer stripping method was constructed to improve the problem of more misclassified pixels. The results show that the layer-by-layer stripping method can recognize more than 77% of the target pixels in the hangar, and the misclassification pixels are reduced by more than 85% compared with the conventional method. It can effectively remove the interference of most other objects in the image, and has higher recognition accuracy and efficiency.