MATLAB to Point Cloud Library

matpcl is pure MATLAB code that allows interfacing with the Point Cloud Library (PCL) tools

Description

matpcl is pure MATLAB code that allows interfacing with the Point Cloud Library (PCL) tools by reading and writing PCD format files. Being pure MATLAB avoids all kinds of headaches in trying to link PCL code into MEX files which involves various grief such as versions of compilers and support libraries such as boost.

There are only four user-level functions:

savepcd() writes a matrix as an optionally coloured point cloud in an ASCII PCD format file.

loadpcd() reads an optionally colored point cloud from a PCD format file (ASCII or binary) and returns a matrix.

pclviewer() writes a matrix to a temporary file and invokes the pcl_viewer app for visualization. This is much much faster for rotating a large point cloud than using a MATLAB 3D plot.

lscpd() shows the attributes of the PCD files in the current directory

Point clouds are considered to be either:

2-d matrices, with one column per point. The rows are X, Y, Z and for a colored point cloud X, Y, Z, R, G, B. (R,G,B) are in the range 0 to 1.

3-d matrices, with planes X, Y, Z and for a colored point cloud X, Y, Z, R, G, B. (R,G,B) are in the range 0 to 1.

Limitations (for now):

- No support for reading/writing normals

• Required Products: MATLAB
• MATLAB release: MATLAB 8.0 (R2012b)
• Other requirements: To run the viewer function pclviewer() you require an install of PCL. Edit pclviewer() to include the path to where your PCL binaries are installed, currently setup for MacOSX.

[Siddharth Choudhary]

Hi Peter,

I get the error:

Index exceeds matrix dimensions.

Error in lzfd (line 59)
out(op:op+len-1) = out(ref:ref+len-1);

Error in loadpcd (line 188)
uncompressed_data = lzfd(compressed_data);

when I run loadpcd on a PCD file link: http://cogrob-dell-desktop.cc.gt.atl.ga.us/test.pcd

Can you check it?

Thanks,

[Peter Corke]

Lauri, sorry to hear this is not working. I don't fully understand the lzfd logic, just transliterated it from another language and it worked ok on the compressed files I had. Can you share the file that caused the problem?

[Lauri]

Hi,

lzfd.m (run by loadpcd.m) gives this error:

Index exceeds matrix dimensions.
Error in lzfd (line 59)
out(op:op+len-1) = out(ref:ref+len-1);

with this data (compressed binary): http://www.cs.tut.fi/~virtanel/skene2.pcd

Unfortunately I have no other datafiles to test nor time to debug lzfd myself.

[Daniel Wolf]

Thanks a lot for your helpful work!
I think I've found a small bug in the loadpcd function processing RGB values. When you split the 4th field into RGB or RGBA values, everytime you shift the field value in the wrong direction using bitshift. Instead of bitshift(rgb, 16) it should be bitshift(rgb, -16) for a right shift instead of a left shift.

[Peter Corke]

Thanks for the addition. regexp(... 'split') is a useful thing, I was looking for the equivalent of python's split() but didn't find this.

binary_compressed format next?

[Will]

For the binary case, I've written some code to handle the rgb / rgba case. Please feel free to use / modify as needed. I hope this helps.

case 'binary'
startPos_fp = ftell(fp);
numFields = numel(siz);
typeTokens = regexp(type,'\s+','split');
fieldTokens = regexp(fields,'\s+','split');

% Just initialize the xyz portion for now
points = zeros(3, npoints);

for i=1:numFields
% map IUF -> int, uint, float
typeToken = typeTokens{i};
switch typeToken
case 'I'
fmt = 'int';
case 'U'
fmt = 'uint';
case 'F'
fmt = 'float';
end

format = ['*' fmt num2str(siz(i)*8)];
fseek(fp, startPos_fp + sum(siz(1:i-1)), 'bof');
data = fread(fp, [1 npoints], format, sum(siz)-siz(i));

fieldToken = fieldTokens{i};
switch fieldToken
case 'x'
points(1,:) = data;
case 'y'
points(2,:) = data;
case 'z'
points(3,:) = data;
case 'rgb'
rgb = typecast(data, 'uint32');
R = double(bitand(255, bitshift(rgb, 16))) /255;
G = double(bitand(255, bitshift(rgb, 8))) /255;
B = double(bitand(255, rgb)) /255;
points = [points(1:3,:); R; G; B];
case 'rgba'
rgb = typecast(data, 'uint32');
R = double(bitand(255, bitshift(rgb, 24))) /255;
G = double(bitand(255, bitshift(rgb, 16))) /255;
B = double(bitand(255, bitshift(rgb, 8))) /255;
A = double(bitand(255, rgb)) /255;
points = [points(1:3,:); R; G; B];
end
end

[Peter Corke]

Thanks for pointing this out, I fixed the problem and uploaded a revised version of the code.

[Albert]

First of all, thank you very much for the software. It is being very useful to me.

In addition, I want to comment you something. I've had a problem trying to run the software. Concretely, in loadpcd.m I was having an error related to the function "numcols" at the line 81. Matlab interpreter was complaining about it in such a way: "undefined function".

Well, I don't like to do this at all, but I changed "numcols(size)" by simply "length(size)" and I think it's working now.

Maybe the error is given by the fact that I'm in Matlab 2012a. As I see, you specified 2012b as your release.

Anyway, I just want to warn you.