虽然用pcl比较多,但是pcl运行比较慢,我很多时候还是喜欢用matlab做一个算法的验证;
项目主要是做一个物体的分割处理。
1.对物体做去噪处理
-
clc
;
-
clear
;
-
B=pcread('test(
1).ply')
;
-
figure(
1)
;
-
pcshow(
B)
;
-
C=pcdenoise(
B,'NumNeighbors',
90,'Threshold',
1)
;
-
figure(
2)
;
-
pcshow(
C)
;
效果如下:(主要就是一个基于最近邻+距离的去噪,类似pcl的统计滤波)
2.做一个roi提取
-
roi = [-inf,+inf
;-inf,inf;-10,inf];
-
indices = findPointsInROI(C, roi)
;
-
-
ptCloudC = select(C,indices)
;
-
figure(
3)
;
-
pcshow(ptCloudC)
;
-
-
roi1 = [-inf,-
8
;-inf,inf;-10,inf];
-
indices1 = findPointsInROI(ptCloudC, roi1)
;
-
ptCloudB = select(ptCloudC,indices1)
;
-
figure(
4)
;
-
pcshow(ptCloudB)
;
3.想要得到上面的包,在做一个平面分割
-
maxDistance =
0.9;
-
referenceVector =
[0,0,-1];
-
maxAngularDistance =
7;
-
[model1,inlierIndices,outlierIndices] =
pcfitplane(ptCloudB,...
-
maxDistance,referenceVector,maxAngularDistance);
-
plane1 =
select(ptCloudB,inlierIndices);
-
remainPtCloud =
select(ptCloudB,outlierIndices);
-
figure
-
pcshow(plane1)
-
title('First
Plane')
-
figure
-
pcshow(remainPtCloud)
-
title('remainPtCloud
Plane')
4.再做roi+滤波
5.聚类
-
minDistance =
0.9;
-
[labels,numClusters] = pcsegdist(E,minDistance);
-
pcshow(E.Location,labels)
-
colormap
(hsv(numClusters))
-
title
('Point Cloud Clusters')
-
disp
(" the number of an jian:")
-
disp
(numClusters)
效果不是特别好,存在一些问题。先验证一下思路是可行的。
5.长宽高的提取
都在matlab 的文件家中。不想复制了。。。