main.asv

%%%单特征波段选择
clc
clear all
ncalSet_ab=xlsread('xunlian.xlsx');
nvalSet_c=xlsread('yuce.xlsx');
load X
load Y
%%%%归一化处理；前14个波段噪声大，不用
X=X(15:233,:)';Y=Y(15:233,:)';
for i=1:960
    for j=1:219
        X0(i,j)=X(i,j)/sqrt(sum(X(i,:).^2));
        Y0(i,j)=Y(i,j)/sqrt(sum(Y(i,:).^2));
    end
end
%%%%划分训练测试
 [mcalSet_ab_shang,mvalSet_c_shang,mcalSet_ab_mean,mvalSet_c_mean]=huafen(X0,Y0,1);
xblock1=[Y0];
[a1,a2]=size(xblock1);
samplemeanA=mean(xblock1); % 平均图片，1 × N
%%%%%%%对样本进行处理；去均值
xmeanA=[];
for i=1:a1
    xmeanA(i,:)=xblock1(i,:)-samplemeanA; % xmean是一个M × N矩阵，xmean每一行保存的数据是“每个图片数据-平均图片”
end;
%%%%协方差矩阵
sigmA=cov(xmeanA);
%%%%斜偏度张量
UA={xblock1',xblock1',xblock1'};
TA = cpdgen(UA);
w=zeros(1,233);
for ii=1:233
    w(ii)=ii;
end
w=w(:,15:233);
k=a2;
%%%%%波段选择主程序
while k>6
    for ii=1:k
        T1=TA;sigma=sigmA;
        T1(ii,:,:)=[];%%%%每次删除ii行
        T1(:,ii,:)=[];
        T1(:,:,ii)=[];
        sigma(ii,:)=[];
        sigma(:,ii)=[];
        R=reshape(T1,(k-1)*(k-1),(k-1));
        R1(:,ii)=svd(R);K(:,ii)=svd(sigma);%%%%高阶奇异值分解
    end
    for j1=1:k
        for i1=1:(k-1)
            s=(R1(i1,j1)/((K(i1,j1)^3)));
        end
        Q(j1)=prod(s,2);
    end
    [n1 n2]=max(Q);%%%删除最大偏度对应波段
    clear Q;   clear K;   clear R;   clear R1;
    TA(n2,:,:)=[];
    TA(:,n2,:)=[];
    TA(:,:,n2)=[];
    sigmA(:,n2)=[];
    sigmA(n2,:)=[];
    w(:,n2)=[];
    [f,k]=size(w);
end

index=[23 25 55 76 91 108 206 210 57 65 177 185 187 190 191];

% index=[59,60,64,65,67,71,73,75,89,94,121,127,132,164,173];%融合
Cal_sel=mcalSet_ab_mean;
Cvl_sel=mvalSet_c_mean;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
samplemean=mean(Cal_sel); % 平均图片，1 × N
xmean1=[];
for i=1:240
    xmean1(i,:)=Cal_sel(i,:)-samplemean; % xmean是一个M × N矩阵，xmean每一行保存的数据是“每个图片数据-平均图片”
end;
sigm=cov(xmean1);   % M * M 阶矩阵
[E1,D1] = eig(sigm);
%——————————计算白化矩阵———————————
whiteningMatrix1 =inv(sqrt(D1)) * E1';
% dewhiteningMatrix = E * sqrt(D);
y=Cal_sel*whiteningMatrix1';
yy=Cvl_sel*whiteningMatrix1';
%%%%%%%%%%% SVM  %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
[bestacc,bestc,bestg] = SVMcgForClass(ncalSet_ab,y,-10,10,-10,10);
% 打印粗略选择结果
str = sprintf( ' Accuracy = %g%% Best c = %g Best g = %g',bestacc,bestc,bestg);
disp(str);
% 利用最佳的参数进行SVM网络训练
cmd = ['-c ',num2str(bestc),' -g ',num2str(bestg)];
model = svmtrain(ncalSet_ab,y, cmd);
% SVM网络训练
[predict_label1,accuracy] = svmpredict(ncalSet_ab,y,model);
Percent1=accuracy(1)/100;
% SVM网络预测
[predict_label2,accuracy] = svmpredict(nvalSet_c,yy,model);
Percent2=accuracy(1)/100;