ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question

How do you take LaserScan data and convert to a point cloud and also do ICP in c++?

asked 2016-06-29 23:17:47 -0500

iopollycron gravatar image

So I am still learning ROS, c++ and Ubuntu (14.04), and I am stuck and haven't been able to implement any similar solution thus far. I am hoping someone else has been in a similar situation and can help me through it. My goal is to take a Hokuyo UST-20LX laser scanner using the urg_node ROS package and do this: Start it up, take the mean of the first 40 scans and set that as my "origin" point for the world. Move the sensor, take that next scan, use an interative closest point algorithm (ICP) to calculate the translation and rotation between the two scans, move to the next point and take another scan and repeat the process indefinitaly. At each step I want to publish the transformation matrix (or new estimated position/velocity) resulting from the ICP step. So far I am still stuck on either working on the range data in the LaserScan message as an entire array to do any manipulation manually, or to use the PCL to work with point clouds.

The only thing I have working is the ROS c++ subscriber code below, I can do a for loop to output each individual element of the range array and convert that into cartesian coordinates but this is so time intensive that it makes no sense to do it this way. As a side note I am unable to verify that a pointCloud2 is published via rviz so I have not been able to see if my PointCloud2 is actually being generated. Can someone shine a light on a more appropriate way or maybe explain what I am doing wrong.

Code so far::

#include "ros/ros.h" #include "std_msgs/String.h" #include "sensor_msgs/LaserScan.h" #include "laser_geometry/laser_geometry.h" #include "tf/message_filter.h" #include "message_filters/subscriber.h" #include "sensor_msgs/PointCloud.h"

void scanCallback(const sensor_msgs::LaserScan::ConstPtr& scan)


    for(int i=0;i<scan->ranges.size();i++)

    {   float range = scan->ranges[i];


//tried converting ranges to point cloud

sensor_msgs::PointCloud2 cloud;

laser_geometry::LaserProjection projector;

    projector.projectLaser(*scan, cloud);

//ROS_INFO("point cloud data =%f",cloud); //tried converting each range to x,y,z coords

float th = scan->angle_min+((float)i/4)*(3.141592653589/180);

float x = range*sin(th);

float y = range*cos(th);

float z = 0;

ROS_INFO("PC Coord = %f, %f, %f, %f",x,y,z,th); 



int main(int argc, char **argv)


ros::init(argc, argv, "subs");

ros::NodeHandle n;

ros::Subscriber subs = n.subscribe("scan", 10, scanCallback);


return 0;


edit retag flag offensive close merge delete

1 Answer

Sort by ยป oldest newest most voted

answered 2016-06-30 08:49:51 -0500

You could use laser_assembler to convert the laser scans into Pointcloud2 messages. It holds the point clouds in a rolling buffer until called by a service. You can either leave it like this or edit the periodic_snapshotter program in the examples folder on github to call the service and publish the pointclouds to a topic at a set rate.

edit flag offensive delete link more

Question Tools

1 follower


Asked: 2016-06-29 23:17:47 -0500

Seen: 5,819 times

Last updated: Jun 30 '16