How do I convert a mesh to a pointcloud for use with pcl ICP

edit

I want to find partial overlaps between meshes in the household object database and pointcloud sensor data. My method is to convert the arm_navigation_msgs::Shape to a pcl::PointCloud and then find clusters that overlap through pcl::IterativeClosestPoint. I have all pieces working, I just think my function for converting the shape meshes to pointclouds is inefficient. Is there a way to take advantage of the underlying structure to speed up the conversion?

This is my current implementation

pcl::PointCloud<pcl::PointXYZ> PCLMeshOverlap::convertMeshToPointcloud(arm_navigation_msgs::Shape mesh)
{
    ROS_INFO("Converting mesh to pointcloud");
    pcl::PointCloud<pcl::PointXYZ> cloud_out;
    pcl::PointXYZ point;
    for (int i = 0; i < mesh.vertices.size(); i++)
    {
        float isNew = true;
        for (int j = 0; j < cloud_out.points.size(); j++)
        {
            if (    mesh.vertices.at(i).x == cloud_out.points.at(j).x &&
                    mesh.vertices.at(i).y == cloud_out.points.at(j).y &&
                    mesh.vertices.at(i).z == cloud_out.points.at(j).z   )
            {
                isNew = false;
                break;
            }
        }
        if (isNew)
        {
            point.x = mesh.vertices.at(i).x;
            point.y = mesh.vertices.at(i).y;
            point.z = mesh.vertices.at(i).z;
            cloud_out.points.push_back(point);
        }
    }
    cloud_out.width = cloud_out.points.size();
    cloud_out.height = 1;
    cloud_out.is_dense = true;
    cloud_out.header.frame_id = "/base_link";
    return cloud_out;
}

*Edited because the original code had some typos