论文部分内容阅读
提出了一种面向目标探测的高光谱图像波段选择方法—波段最大筛选法(MBS),它将每个波段的图像看成一条波段向量,以两个最不相似的波段作为初始波段,每次从剩余波段中选取一个和已选波段最不相似的波段,通过对波段相似性阈值的合理调节,保证了目标探测算子在所选波段上探测效果最佳。为了验证MBS的有效性,对机载可见光/红外成像光谱仪(AVIRIS)获取的两幅真实高光谱图像进行了实验,结果表明,MBS选出的波段分别占据全部波段的15%和9%,从而使目标探测算子ACE和AMF在其上的探测性能有了明显改善。
This paper proposes a target-detection-based hyperspectral image band selection method (MBS), which takes the image of each band as a band vector and takes the two most dissimilar bands as the initial band, each time By selecting a band which is the most dissimilar to the selected band from the remaining bands, the reasonable detection of the target detection operator in the selected band is ensured by reasonably adjusting the band similarity threshold. In order to verify the effectiveness of MBS, two real hyperspectral images acquired by AVIRIS were tested. The results show that the selected bands of MBS occupy 15% and 9% of the total band, respectively The performance of target detection operator ACE and AMF has been significantly improved.