MATLAB: How to streamline this code to average certain rows

loopMATLAB

I have a three row column data: lon lat and Z. What I want to do is to round longitude and latitude to 2 digits. For example, 65.3796 will be rounded 65.38. Then the program will sort the data based on longitude and latitude, and average any rows with identical longitudes and latitudes. For example,
The below matrix:
65.39 23.56  10.5
66.70 25.36  6.7
66.70 25.36  7.8
Will become:
65.39 23.56  10.5
66.70 25.36  7.25
Here is my code. The problem is that I have several dozens of millions of data. It will take forever to complete. Anyone could help me create a more streamlined code, so that the program will be much faster?
I really appreciate your help!
clear all
close all
clc
load data/data_3row % D: lon lat Z
lon = D(:,1);
lat = D(:,2);
for i=1:length(lon)
    lon(i,1) = round(lon(i,1)*100)/100;
    lat(i,1) = round(lat(i,1)*100)/100;
end
D2 = [lon, lat, D(:,3)];
D3 = sortrows(D2, [1:2]);
D3 = [D3; 0 0 0];
[m n] = size(D3);
Ind1 = D3(1,1);
Ind2 = D3(1,2);
row  = D3(1,:);
D4 = [];
n4 = 0;
for i=2:m
      if D3(i,1)==Ind1 & D3(i,2)==Ind2
          row = [row; D3(i,:)];
      else
          n4 = n4+1;
          D4(n4,:) = [row(1,1), row(1,2), nanmean(row(:,3))];
          Ind1 = D3(i,1);
          Ind2 = D3(i,2);
          row  = D3(i,:);
      end
  end
save('data/data_3row_avg.mat', 'D4');

Best Answer

Replace
for i=1:length(lon)
    lon(i,1) = round(lon(i,1)*100)/100;
    lat(i,1) = round(lat(i,1)*100)/100;
end
with
lon = round(lon, 2);
lat = round(lat, 2);
After that,
vals = D(:,3);
[ulon, ~, ulonidx] = unique(lon);
[ulat, ~, ulatidx] = unique(lat);
meanarray = accumarray([ulonidx, ulatidx], vals, [], @mean, nan);
valididx = find( ~isnan(meanarray) );
[lonidx, latidx] = ind2sub( size(meanarray), valididx );
means = meanarray(valididx);
D4 = [ulon(lonidx), ulat(latidx), means];
Related Question