What is the Stereocam Field of View (FOV)?

[mabitter-x]

### Description Hello! My questions are relating to the top-view Stereocams, specifically relating to their calibration and general operation. Forgive me if this information is somewhere in the documentation, however I have done an extensive search and not found any clarity on a few issues. 1) Is the FOV set throughout the season? As the height of the gantry is changing as the plants grow, is the FOV changing? I've gotten a few different readings between the git issues, code-base, the metadata, the lemnatec documentation, and what i'm seeing based on my trigonometry... and thought maybe you guys may have landed on a number or system you are happy with now? 2) What calibration information and data samples might you be able to provide? I've looked at all the specifications that are available through the databases, however think there must be more specific data and test done with color correction, determining the camera nodal points, etc. Was there any imaging of scale references (like a yardstick) from different heights, or with other fiducial markers? Is there any more detailed documentation of these processes? 3) Camera baseline (intra-axial distance) - I couldn't find this value, or much other than the offset from box corner to some centerpoint of the stereo rig. This value also differed slightly in the lemnatec documentation and the metadata. Assuming the orientation (rx, ry, rz) values are near-zero? 4) I remember someone mentioning the distance to the ground plane gets a little further from the plane of the gantry as you move north in the field, is there any documentation of this anywhere? Actual measures? I suppose some LIDAR data from early season before the plants have grown could be used... 5) Has the calibration or FOV protocols changed since Season2? If so, what should we know about using Season2 data? Thanks for any thoughts and data you might be able to provide! Thanks! Matthew ### Completion Criteria * [ ] Provide Stereocam FOV, any protocols and any relevant calibration information and datasets across seasons 2, 3, and the current season.

[dlebauer]

I'll take a stab at 1 and 4.

@pless, @smarshall-bmr, @solmazhajmohammadi could you please respond to questions 2. 3, and 5?

@craig-willis and / or @max-zilla can you please confirm my assertions about the metadata?

Question 1

Is the FOV set throughout the season?

The FOV should be fixed relative to a plane 2m below the camera

As the height of the gantry is changing as the plants grow, is the FOV changing?

Relative to the ground, the FOV is changing. But metadata provides FOV relative to 2m from the camera

I've gotten a few different readings between the git issues, code-base, the metadata, the lemnatec documentation,

Since the issue first came up (https://github.com/terraref/computing-pipeline/issues/126) I think we have determined the correct FOV, and the canonical source of fixed sensor metadata is under the 'metadata' tab here: https://terraref.ncsa.illinois.edu/clowder/datasets/5873a8ae4f0cad7d8131ac0e?space=594bda394f0cf8bbab271dfb (see: field of view at 2m in X- Y- direction [m]: [1.857 1.246])

We only use a few pieces of the variable metadata from the .json metadata files that are generated with each data set (in the case of the stereo RGB cameras, each pair of images). We do not rely on the documentation or the fixed metadata that is contained in those files, since they have historically not been reliable (e.g. because they haven't been updated).

We will update all of the file-specific metadata in Clowder with the canonical fixed metadata when we reprocess the data in September prior to the beta release in November. Until then, there may be errors in these datasets.

and what i'm seeing based on my trigonometry...

Please let us know if your trigonometry differs from our metadata!

Question 4:

the distance to the ground plane gets a little further from the plane of the gantry as you move north in the field, is there any documentation of this anywhere?

No, this isn't currently documented but perhaps should be!

@Andrade-Pedro took an elevation transect before the gantry was built on July 30 2015:

We did a 3D scan about a week after planting (April 28 2017 + almost daily through emergence that should provide a good baseline. @smarshall-bmr did you do bare-ground scans as well?

[smarshall-bmr]

Question 1

The field of view is fixed at 28.4 degrees on the x and 21.2 degrees on the y axis for each camera. This equates to 101.5x74.9cm at 2m. @dlebauer we need to make sure this value gets updated everywhere since the original values given by LemnaTec are still floating around.

Question 2

There have been many calibration images taken, many of which were used in the LemnaTec calibration report. There are images for resolution, color correction, and motion blur. I have done "unofficial" checks for chromatic aberration and DoF.

Question 3

Intra-axial distance on the two cameras is 30.0cm. There are multiple values reported for the offset because different people have taken different points as the instrument location. I'll have to confirm, but I believe that the metadata location is a point halfway between the base of the two camera lenses. (I'll have to confirm this)

Question 4

The field is laser leveled between plantings to within 2cm. There is an initial 3D scan taken of the field before emergence that will provide the most accurate measures.

Question 5

The aperture and focus were finalized midway through season 2, the change should be reflected in the metadata. The cameras were checked before season 4 (the current season) and the focus had drifted slightly and was corrected.

[craig-willis]

Per team meeting, there are two additional questions for (me and @max-zilla) a) is the sensor fixed metadata in Clowder the right place to look for the FOV data b) has it been applied to the data (@max-zilla )

[craig-willis]

@dlebauer @max-zilla Per @smarshall-bmr comment, the values in the fixed metadata in Clowder are wrong. This was apparently all covered in https://github.com/terraref/computing-pipeline/issues/126. It looks like the correct values were likely harcoded in the extractor in question, but I didn't see an issue to update these values in the fixed metadata.

I've created a PR for the fixed metadata. Once approved/merged, I'll push the change to clowder: https://github.com/terraref/sensor-metadata/pull/6

[craig-willis]

FOV data updated in Clowder https://terraref.ncsa.illinois.edu/clowder/datasets/5873a8ae4f0cad7d8131ac0e

[ghost]

@mabitter-x do you have any further questions or can this be closed?

[mabitter-x]

Thank you all for this wealth of information, looks good to me!

On Thu, Jul 27, 2017 at 9:09 AM, Rachel Shekar notifications@github.com wrote:

@mabitter-x https://github.com/mabitter-x do you have any further questions or can this be closed?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/terraref/computing-pipeline/issues/341#issuecomment-318409546, or mute the thread https://github.com/notifications/unsubscribe-auth/AZovWVK9tubTNaL3VjWP8xtN3qwj6NIDks5sSLY_gaJpZM4OYj_8 .

--

Matthew Bitterman Senior Design Engineer 919-649-0337 <(919)%20649-0337> solveforx.com