基于图像对比度的监督波段选择方法  

A Supervised Band Selection Method Based on Imag Contrast

作　　者：刘恂[1] 华文深[1] 杨佳[1] 刘晓光[1] 

机构地区：[1]军械工程学院,石家庄050003

出　　处：《激光杂志》2015年第2期27-29,34,共4页Laser Journal

摘　　要：约束能量最小化(Constrained Energy Minimization,CEM)是一种常见的小目标探测算法,为提高该算法的探测效率,提出了基于图像对比度的监督波段选择方法(CBS)。该方法针对CEM探测原理,依据高光谱数据各波段图像的对比度以及波段间相关性,选择出对比度大且波段间相关性小的波段组合用于小目标探测。实验发现基于CBS探测结果明显优于基于图像方差的波段选择方法(VBS)的探测结果。在得到质量相当的探测图像前提下,使用CBS需要的波段数远小于使用VBS需要的波段数。CBS对于降低CEM探测的计算量,提高探测的实时性具有重要意义。A supervised band selection approach based on image contrast（ CBS） is proposed. The proposed method aims at increasing the efficiency of constrained energy minimization（ CEM）,which is detector widely used for the detection of small targets. Image contrast and correlation of inter-band are taken into consideration to select a band subset which is characterized by high image contrast and low correlation. Through experiments,it has been demonstrated that detecting result based on CBS outperforms the method using variance based band selection（ VBS）. To achieve similar result,the number of bands need when CBS is utilized is roughly lower than that when VBS is used. Moreover,the detecting result of using 20 bands selected by CBS approximate the result when whole bands（ 89 bands） are employed. The proposed method is meaningful to decreasing the time need for detection and also dose good for increasing the detecting efficiency.

关 键 词：遥感 高光谱 波段选择 小目标探测 对比度 

分 类 号：TP911[自动化与计算机技术]