1 简介
针对人工检测金属表面缺陷效率低,主观意识强,无法长时间工作等缺点,提出一种基于支持向量机监督检测,分类以及测量的金属表面缺陷的方法,并使用Matlab软件设计一个图形用户界面(GUI),便于检测人员使用.研究中先对工厂采集的图像进行Gabor滤波和对比度增强的前处理.然后使用方向梯度直方图(HOG)和灰度共生矩阵(GLCM)进行特征提取,为后续提高检测准确率打下基础.最后对于每种缺陷类型都采用270张图片进行模型训练,利用训练好的模型对测试图片进行测试.实验结果表明,最终3种缺陷类型的总正确率为88.9%,该检测方法能够有效地检测出金属表面缺陷.
2 部分代码
c2='Sc_72';
c1='.\NEU surface defect database\';
c3='.bmp';
c=[c1,c2,c3];
a6=imread(c);
ma=max(max(a6));
mi=min(min(a6));
for ii=1:200
for jj=1:200
a6(ii,jj)=255*double(a6(ii,jj)-mi)/double(ma-mi);
end
end
figure(1),imshow(a6,[]);
a6=adapthisteq(a6);
figure(1),imshow(a6,[]);
Gaborfiter=Ga(1.57,10);
aaa=filter2(Gaborfiter,a6,'valid');
%figure(1),imshow(aaa,[]);
%aa=imgaborfilt(a5,10,0);
%bb=imgaborfilt(a2,10,90);
bb=Ga(0,25);
bbb=filter2(bb,a6,'valid');
cc=Ga(0.785,25);
ccc=filter2(cc,a6,'valid');
dd=Ga(2.36,25);
ddd=filter2(dd,a6,'valid');
%figure(2),imshow(aa,[])
h=fspecial('average',5);
ab=filter2(h,abs(abs(aaa)+abs(ddd)-abs(bbb)-abs(ccc)));
ab=filter2(h,ab);
ab=uint8(fix(255*ab./max(max(ab))));
%ab=im2bw(ab,graythresh(ab));
aaa=uint8(fix(255*abs(aaa)./max(max(abs(aaa)))));
ss=thres(ab);
ab1=im2bw(ab,ss/255);
%figure(2),imshow(a6,[]);
%[bw,thres]=edge(ab,'log');
% s=strel('disk',2);
% ab=imerode(ab,s);
[B,L]=bwboundaries(ab1);
[x,y]=size(B);
mm=0;
for i=1:x
ll=length(B{i});
if ll>mm
mm=ll;
end
end
if mm<110
fprintf('缺陷是点\n');
xi=zeros(x,1);
yi=zeros(x,1);
xx=zeros(x,1);
yy=zeros(x,1);
xc=zeros(x,1);
yc=zeros(x,1);
lll=zeros(x,1);
for i=1:x
lll(i)=length(B{i});
ll=length(B{i});
if ll>10
for ij=1:ll
xi(i)=xi(i)+double(ab(B{i}(ij,1),B{i}(ij,2)));
yi(i)=yi(i)+double(ab(B{i}(ij,1),B{i}(ij,2)));
xx(i)=xx(i)+double(ab(B{i}(ij,1),B{i}(ij,2))).*B{i}(ij,1);
yy(i)=yy(i)+double(ab(B{i}(ij,1),B{i}(ij,2))).*B{i}(ij,2);
end
end
xc(i)=xx(i)./xi(i);
yc(i)=yy(i)./yi(i);
end
xc=xc(~isnan(xc));
yc=yc(~isnan(yc));
figure(4),imshow(a6);
hold on;
plot(yc+12,xc+12,'*');
[dd,bb]=size(xc);
fprintf('一共有%d以上个点',dd);
else
%ab1=bwmorph(ab1,'thin',5);
% ss=thres(a6);
% ab1=im2bw(a6,ss/255);
figure(3),imshow(ab1);
a6(1:12,:)=[];
a6(177:188,:)=[];
a6(:,1:12)=[];
a6(:,177:188)=[];
mu1=uint8(ab1).*a6;
mu2=uint8(1-ab1).*a6;
t0=mu1(mu1~=0);
m1=mean(t0);
t1=mu2(mu2~=0);
m2=mean(t1);
MAPPING=getmapping(16,'riu2');
%lbps=lbp(a6,2,16,MAPPING,'hist');
%predict_label = svmclassify(svmModel,lbps);
if m1<m2+20
fprintf('缺陷是斑点\n');
figure(5),imshow(mu1);
all=bwarea(ab1);
pro=all/176/176;
fprintf('斑点所占总比为%.3f',pro);
figure(4),imshow(a6);
hold on;
for k = 1:length(B)
boundary = B{k};
plot(boundary(:,2), boundary(:,1), 'r', 'LineWidth', 2)
end
else
ss=thres(a6);
ab1=im2bw(a6,ss/255);
%ab1=im2bw(a6,graythresh(a6));
ab1=imclose(ab1,strel('square',5));
ab1=bwmorph(ab1,'thin',5);
%[B,L]=bwboundaries(ab1);
ab1=edge(ab1,'Canny');
figure(5),imshow(ab1);
hold on;
fprintf('缺陷是划痕\n');
[H,T,R] = hough(ab1,'RhoResolution',5,'Theta',-90:0.1:89.9);
P=houghpeaks(H,35,'threshold',ceil(0.3*max(H(:))));
lines = houghlines(ab1,T,R,P,'FillGap',5,'MinLength',7);
point=zeros(length(lines),4);
ang=zeros(length(lines),1);
for k = 1:length(lines)
xy = [lines(k).point1; lines(k).point2];
plot(xy(:,1),xy(:,2),'LineWidth',2,'Color','green');
% Plot beginnings and ends of lines
plot(xy(1,1),xy(1,2),'x','LineWidth',2,'Color','yellow');
plot(xy(2,1),xy(2,2),'x','LineWidth',2,'Color','red');
dis=sqrt((xy(1,1)-xy(2,1))^2+(xy(1,2)-xy(2,2))^2);
if dis>30
point(k,1:2)=lines(k).point1;
point(k,3:4)=lines(k).point2;
ang(k)=lines(k).theta;
end
end
point(any(point==0,2),:)=[];
ang(any(ang==0,2),:)=[];
cen=zeros(length(ang),2);
po=zeros(length(ang),4);
an=zeros(length(ang),1);
p1=zeros(length(ang),4);
an1=zeros(length(ang),1);
thm=mean(abs(ang));
if abs(thm)<45
thm=40;
else
thm=60;
end
p1(all(p1==0,2),:)=[];
an1(all(an1==0,2),:)=[];
le=length(cen);
for kk=1:le
cen(kk,1)=(point(kk,1)+point(kk,3))/2;
cen(kk,2)=(point(kk,2)+point(kk,4))/2;
point(kk,5)=cen(kk,1);
point(kk,6)=cen(kk,2);
end
if thm==40
[point,ind]=sortrows(point,5);
end
if thm==60
point=sortrows(point,6);
end
anm=0;
cend=zeros(length(cen)-2,1);
figure(6),imshow(a6);
hold on;
count=0;
if thm==60
for kk=2:le
if abs(point(kk-1,6)-point(kk,6))<5
po(kk,:)=point(kk,1:4);
an(kk)=ang(kk);
anm=mean(an(1:kk));
else
count=count+1;
po(all(po==0,2),:)=[];
max_x=max(max(po(:,1)),max(po(:,3)));
min_x=min(min(po(:,1)),min(po(:,3)));
max_y=max(max(po(:,2)),max(po(:,4)));
min_y=min(min(po(:,2)),min(po(:,4)));
if anm>0
plot(max_x,max_y,'x','LineWidth',2,'Color','green');
plot(min_x,min_y,'x','LineWidth',2,'Color','blue');
xy(1,:)=[max_x,max_y];
xy(2,:)=[min_x,min_y];
plot(xy(:,1),xy(:,2),'LineWidth',2,'Color','green');
else
plot(max_x,min_y,'x','LineWidth',2,'Color','green');
plot(min_x,max_y,'x','LineWidth',2,'Color','blue');
xy(1,:)=[max_x,min_y];
xy(2,:)=[min_x,max_y];
plot(xy(:,1),xy(:,2),'LineWidth',2,'Color','green');
end
dis=sqrt((max_x-min_x)^2+(max_y-min_y)^2);
fprintf('第%d条划痕长%.2f\n',count,dis);
po=zeros(length(ang),4);
end
end
po(all(po==0,2),:)=[];
max_x=max(max(po(:,1)),max(po(:,3)));
min_x=min(min(po(:,1)),min(po(:,3)));
max_y=max(max(po(:,2)),max(po(:,4)));
min_y=min(min(po(:,2)),min(po(:,4)));
if anm>0
plot(max_x,max_y,'x','LineWidth',2,'Color','green');
plot(min_x,min_y,'x','LineWidth',2,'Color','blue');
xy(1,:)=[max_x,max_y];
xy(2,:)=[min_x,min_y];
plot(xy(:,1),xy(:,2),'LineWidth',2,'Color','green');
else
plot(max_x,min_y,'x','LineWidth',2,'Color','green');
plot(min_x,max_y,'x','LineWidth',2,'Color','blue');
xy(1,:)=[max_x,min_y];
xy(2,:)=[min_x,max_y];
plot(xy(:,1),xy(:,2),'LineWidth',2,'Color','green');
end
count=count+1;
dis=sqrt((max_x-min_x)^2+(max_y-min_y)^2);
fprintf('第%d条划痕长%.2f\n',count,dis);
fprintf('一共有%d条线\n',count);
elseif thm==40
for kk=2:le
if abs(point(kk-1,5)-point(kk,5))<10
po(kk,:)=point(kk,1:4);
an(kk)=ang(kk);
anm=mean(an(1:kk));
%poi=mean(po(kk,1));
%poj=mean(po(kk,3));
else
% an(kk)=ang(kk);
% anm=mean(an(1:kk));
% poi=mean(po(kk,1));
% poj=mean(po(kk,3));
count=count+1;
po(all(po==0,2),:)=[];
max_x=max(max(po(:,1)),max(po(:,3)));
min_x=min(min(po(:,1)),min(po(:,3)));
max_y=max(max(po(:,2)),max(po(:,4)));
min_y=min(min(po(:,2)),min(po(:,4)));
if anm>0
plot(max_x,max_y,'x','LineWidth',2,'Color','green');
plot(min_x,min_y,'x','LineWidth',2,'Color','blue');
xy(1,:)=[max_x,max_y];
xy(2,:)=[min_x,min_y];
plot(xy(:,1),xy(:,2),'LineWidth',2,'Color','green');
else
plot(max_x,min_y,'x','LineWidth',2,'Color','green');
plot(min_x,max_y,'x','LineWidth',2,'Color','blue');
xy(1,:)=[max_x,min_y];
xy(2,:)=[min_x,max_y];
plot(xy(:,1),xy(:,2),'LineWidth',2,'Color','green');
end
dis=sqrt((max_x-min_x)^2+(max_y-min_y)^2);
fprintf('第%d条划痕长%.2f\n',count,dis);
po=zeros(length(ang),4);
end
end
po(all(po==0,2),:)=[];
%poi=mean(po(:,1));
%poj=mean(po(:,3));
max_x=max(max(po(:,1)),max(po(:,3)));
min_x=min(min(po(:,1)),min(po(:,3)));
max_y=max(max(po(:,2)),max(po(:,4)));
min_y=min(min(po(:,2)),min(po(:,4)));
if anm>0
plot(max_x,max_y,'x','LineWidth',2,'Color','green');
plot(min_x,min_y,'x','LineWidth',2,'Color','blue');
xy(1,:)=[max_x,max_y];
xy(2,:)=[min_x,min_y];
plot(xy(:,1),xy(:,2),'LineWidth',2,'Color','green');
else
plot(max_x,min_y,'x','LineWidth',2,'Color','green');
plot(min_x,max_y,'x','LineWidth',2,'Color','blue');
xy(1,:)=[max_x,min_y];
xy(2,:)=[min_x,max_y];
plot(xy(:,1),xy(:,2),'LineWidth',2,'Color','green');
end
count=count+1;
dis=sqrt((max_x-min_x)^2+(max_y-min_y)^2);
fprintf('第%d条划痕长%.2f\n',count,dis);
fprintf('一共有%d条线\n',count);
end
end
end
clear;
3 仿真结果
4 参考文献
[1]张旭等. "一种基于支持向量机的金属表面缺陷检测方法.".
部分理论引用网络文献，若有侵权联系博主删除。
5 MATLAB代码与数据下载地址
见博客主页
