Can't get ground points for depth-image-encoding.

edit

Hello everybody,

I'm trying to convert a depth-image from a stereocamera to a point-cloud like shown in the following code:

void DepthImageCallback(const sensor_msgs::ImageConstPtr& msg)
{
    if (!cam_model.initialized())
        return;

    sensor_msgs::PointCloud2::Ptr cloud_msg(new sensor_msgs::PointCloud2);
    cloud_msg->header = msg->header;

    cloud_msg->header.frame_id = "zed3d";
    cloud_msg->header.stamp = ros::Time::now() + delay;

    cloud_msg->height = msg->height;
    cloud_msg->width  = msg->width;
    cloud_msg->is_dense = false;
    cloud_msg->is_bigendian = false;

    sensor_msgs::PointCloud2Modifier pcd_modifier(*cloud_msg);
    pcd_modifier.setPointCloud2FieldsByString(1, "xyz");

    if (msg->encoding == sensor_msgs::image_encodings::TYPE_16UC1)
    {
    convert<uint16_t>(msg, cloud_msg, cam_model);
    }
    else if (msg->encoding == sensor_msgs::image_encodings::TYPE_32FC1)
    {
    convert<float>(msg, cloud_msg, cam_model);
    }
    else
    {
    ROS_ERROR("Depth image has unsupported encoding");
    return;
    }
    cloud_publisher.publish(cloud_msg);

The converting process is successfull and i get a point-cloud of type PointCloud2.

As my goal is to create a ground-model out of the point-cloud data, I need to see the "ground-points".

That's where the problem is.

Do you have any idea, what I'm doing wrong and how I can achieve to see the ground-points?

Thank you very much in advance!