聚类算法基于matlab划分法k-means聚类算法【含Matlab源码 1941期】.zip

function [ resX,resY, resZ,seedX,seedY,seedZ,record] = FunK_mean3D( x,y,z,k,r) while 1 j = 1; seedX = zeros(1,k); seedY = zeros(1,k); seedZ = zeros(1,k); oldSeedX = zeros(1,k); oldSeedY = zeros(1,k); oldSeedZ = zeros(1,k); resX = zeros(k,length(x)); resY = zeros(k,length(y)); resZ = zeros(k,length(z)); record = zeros(1,k); % 用来记录resX中每一行有效元素的个数 p=2; while 1 if p == 0 break; end for i = 1:k % 产生k个随机种子, 注意： 随机种子是来自元素集合 seedX(i) = x(ceil(rand()*length(resX))); seedY(i) = y(ceil(rand()*length(resY))); seedZ(i) = z(ceil(rand()*length(resZ))); end for j=k:-1:2 p=0; for m=j-1:-1:1 power(seedX(m)-seedX(j),2)+power(seedY(m)-seedY(j),2)+power(seedZ(m)-seedZ(j),2) if (power(seedX(m)-seedX(j),2)+power(seedY(m)-seedY(j),2)+power(seedZ(m)-seedZ(j),2))<r p=1 break; end end if p==1 break; end end end while 1 record(:) = 0; % 重置为零 resX(:) = 0; resY(:) = 0; resZ(:) = 0; for i = 1:length(x) % 对所有元素遍历 % 下面是判断本次元素应该归为哪一类，这里我们是根据欧几里得距离进行类别判定 % k-mean算法认为元素应该归为距离最近的种子代表的类 distanceMin = 1; for j = 2:k if (power(x(i)-seedX(distanceMin),2)+power(y(i)-seedY(distanceMin),2)+power(z(i)-seedZ(distanceMin),2))... > (power(x(i)-seedX(j),2) + power(y(i)-seedY(j),2)+power(z(i)-seedZ(j),2)) distanceMin = j; end end % 将本次元素点进行类别归并 resX(distanceMin,record(distanceMin) +1) = x(i); resY(distanceMin,record(distanceMin) +1) = y(i); resZ(distanceMin,record(distanceMin) +1) = z(i); record(distanceMin) = record(distanceMin) + 1; end oldSeedX = seedX; oldSeedY = seedY; oldSeedZ = seedZ; % 移动种子至其类中心 record; for i = 1:k if record(i) == 0 continue; end seedX(i) = sum(resX(i,:))/record(i); seedY(i) = sum(resY(i,:))/record(i); seedZ(i) = sum(resZ(i,:))/record(i); end maxPos = max(record); resX = resX(:,1:maxPos); resY = resY(:,1:maxPos); resZ = resZ(:,1:maxPos); % 如果本次得到的种子和上次的种子一致，则认为分类完毕。 if mean([seedX == oldSeedX&seedY == oldSeedY&seedZ == oldSeedZ]) == 1 % 这句话所想表达的意思就是 if seedX == oldSeedX && seedY == oldSeedY break; end end record( find(record==0))=[]; if length(record) == k break; end end end