Get pointcloud data from ros messages fail

[HaogeZhou]

Thanks for your work, I am very excited to try it! I am using Ubuntu 20.04 + ROS noetic + Ouster os1-64 lidar with inner imu data + zed2i camera. I was trying to start the r3live_bag_ouster.launch file to do mapping, but when I started to run bag, I received the error message related to the point cloud. See attached below.

Could someone tell me how to fix this error? Thanks in advance.

[alwynmathew]

For some reason, it only works for lidar scans with fields.size() < 3 or fields.size() == 8 && fields[ 3 ].name == "intensity" && fields[ 4 ].name == "normal_x" or fields.size() == 4 && fields[ 3 ].name == "rgb" format as shown below: https://github.com/hku-mars/r3live/blob/6143a38537f28cb36eb24e9bbe2e39c8f7967157/r3live/src/r3live_lio.cpp#L91

[HaogeZhou]

Hi @alwynmathew , I also seen this code. Here is my lidar topic output.

> header: 
>   seq: 4430
>   stamp: 
>     secs: 528
>     nsecs: 948532131
>   frame_id: "os_sensor"
> height: 64
> width: 2048
> fields: 
>   - 
>     name: "x"
>     offset: 0
>     datatype: 7
>     count: 1
>   - 
>     name: "y"
>     offset: 4
>     datatype: 7
>     count: 1
>   - 
>     name: "z"
>     offset: 8
>     datatype: 7
>     count: 1
>   - 
>     name: "intensity"
>     offset: 16
>     datatype: 7
>     count: 1
>   - 
>     name: "t"
>     offset: 20
>     datatype: 6
>     count: 1
>   - 
>     name: "reflectivity"
>     offset: 24
>     datatype: 4
>     count: 1
>   - 
>     name: "ring"
>     offset: 26
>     datatype: 4
>     count: 1
>   - 
>     name: "ambient"
>     offset: 28
>     datatype: 4
>     count: 1
>   - 
>     name: "range"
>     offset: 32
>     datatype: 6
>     count: 1
> is_bigendian: False
> point_step: 48
> row_step: 98304

I have no idea about knowing my lidar(ouster os1-64) field size is suitable or not. How can I get the field size of my lidar?

[alwynmathew]

Your LiDAR field size is 9, as shown in the error screenshot above. A workaround is adding the below statement after L112 in r3live_lio.cpp

else if ( ( msg->fields.size() == 9 ) ) 
        {
            pcl::fromROSMsg( *msg, pcl_pc );
            return true;
        }

This should make the package work, but still, you will get a warning that normals are missing. They expect the input to be in PointXYZINormal format. Good luck!

[HaogeZhou]

Thanks for your help! As you imagine, the point cloud is shown in Rviz, but the "normal_x" "normal_y" "normal_z" are missing. I'm trying to change PointXYZINormal to PointXYZI, but it seems not work.

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[zxhasd]

感谢您的帮助！正如您想象的那样，点云显示在 Rviz 中，但缺少“normal_x”、“normal_y”、“normal_z”。 我正在尝试将 PointXYZINormal 更改为 PointXYZI，但似乎不起作用。

请问您解决这个问题了吗