Oct 29, 2024
Sphaeroptica Manual
- Yann Pollet1
- 1Royal Belgian Institute of Natural Sciences
Protocol Citation: Yann Pollet 2024. Sphaeroptica Manual. protocols.io https://dx.doi.org/10.17504/protocols.io.6qpvr352pvmk/v1
License: This is an open access protocol distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Protocol status: Working
We use this protocol and it's working
Created: February 07, 2024
Last Modified: October 29, 2024
Protocol Integer ID: 94810
Keywords: sphaeroptica manual manual, sphaeroptica, manual, available on github
Abstract
Manual to use Sphaeroptica 1.0, available on GitHub: https://github.com/ypollet/Sphaeroptica
Troubleshooting
Installation
We have been able to run Sphaeroptica 1.0 on a laptop with an Intel i5-7200 and 8GB of RAM on Ubuntu 20.04. It is not necessary to have a Graphic card to run the program, even though it will be slightly faster. It is not heavy at all.
Here is a list of the requirements needed to run Sphaeroptica 1.0 :
- python >= 3.8
- numpy
- scipy
- PySide6
- imutils
- opencv-python
- matplotlib
- pandas
- bs4
- lxml
Make sure that only opencv-python is installed, having other opencv packages installed can create some bugs.
If OpenCV and PySide6 show an error, uninstall opencv-python and reinstall it with this command :
Command
new command name
The easiest way to install the requirements is to install the pre-configured Anaconda environment.
On Ubuntu 22.04:
Command
install required packages on a virtual environment (Ubuntu 22.04)
Install requirements with conda
On Windows :
Command
Install required packages on a virtual environment (Windows)
Install requirements conda
After the initialization of the conda environment, restart the terminal and activate it :
Command
activate your new environment
activate conda environment
But if you prefer to install them through pip, we created a file that contains all the requirements, you can install them with these commands :
Command
install requirements through pip (Ubuntu 22.04)
pip install requirements
How to make your own sphere of images
Sphaeroptica 1.0 has two requirements :
- A set of images taken around an object (at a similar distance and with only one camera)
- the calibration of these images
Image Acquisition
For the first requirement, every technique is allowed, as long as you have enough pictures to have enough overlap for their calibration. As it is very visual, because the virtual camera hops between each picture, it is encouraged to have enough pictures to have a smooth movement on it.
Focus Stacking (optional)
It is necessary for smaller specimens with a lot of details, and focus stacking distortion is negligible for the calibration and triangulation (as our tests suggest and Olkowicz et al., 2019 ).
The best stacker software that we have tested are Zerene Stacker and Helicon Focus (Here) and Helicon Focus uses the GPU for its computations (thus, is much faster).
However, these are softwares with a paid license. A free open-source solution would be Enfuse (even though it gives a halo around the object).
Calibration
For the calibration step, we would recommend you to use softwares that you are already used to. It will result in a faster and better process for you.
You have to simply make sure that the exported data from the software has to be conform to Sphaeroptica. The format needed is shown at Step 9.
In our case, we used Agisoft Metashape Professional Edition, a licensed photogrammetry software. And thus developed converters around the data that we get from Agisoft Metashape.
If you use another software, you will have to convert the data yourself.
However, to make the program more accessible to every user, we are developing a converter from data exported by COLMAP, as it is a free open-source project.
Please, make sure that the pictures are correctly oriented (calibration) and rotated (rotation of the picture).
Sphaeroptica 1.0 does not change these parameters and not doing so could land some funny results.
Frontal view should be at (0,0) with lateral views at (-90,0) (right side) and (90,0) (left side). If needed, rotate the specimen in your SfM software before exporting its calibration parameters.
If you use Agisoft Metashape, here are the steps you have to do to convert your data for Sphaeroptica 1.0.
When your images are calibrated (with “Align Photos”), you can export the intrinsics parameters with Tools > Camera Calibration > Adjusted, and export it in "OpenCV Camera Calibration (*.xml)" format. This will give you the file needed for the intrinsics values in Sphaeroptica 1.0.
For the extrinsics, you have to go to File > Export > Export Cameras, and export the file as "Omega Phi Kappa (*txt)".
This CSV file still has to be converted into the needed JSON file.
Fortunately, we developed a script that does that and that is available at scripts/additional/import_cameras_to_sphaeroptica.py
Command
new command name
Create a Sphaeroptica 1.0 project
Start the application
Command
new command name
Sphaeroptica 1.0 requires that the “Data” folder contains all the images, to have the intrinsic parameters in a XML file, and the extrinsic parameters in a JSON file for these images.
Command
data folder
When you decide to create a new project file, this is the window that will appear :
Screen of project creation
On “Image Directory”, browse to the data folder and select it.
On “Intrinsics”, select the XML file containing the intrinsics values.
The file should look like this :
<?xml version="1.0"?>
<opencv_storage>
<calibration_Time>"Mon Apr 17 14:24:32 2023"</calibration_Time>
<image_Width>width</image_Width>
<image_Height>height</image_Height>
<Camera_Matrix>
<rows>3</rows>
<cols>3</cols>
<dt>d</dt>
<data>
fx 0. cx 0. fy cy 0. 0. 1.</data></Camera_Matrix>
<Distortion_Coefficients>
<rows>n</rows>
<cols>1</cols>
<dt>d</dt>
<data>
k1 k2 p1 p2 ...</data></Distortion_Coefficients>
</opencv_storage>
On “Extrinsics”, select the json file containing the extrinsics values for each of the images The file should look like this :
{
"IMAGE1.jpg": {
"matrix": [
[r11,r12,r13,t1],
[r21,r22,r23,t2],
[r31,r32,r33,t3],
[0.0,0.0,0.0,1.0]
]
},
"IMAGE2.jpg":
...
}
If you have already a folder of thumbnails, you can specify it to Sphaeroptia 1.0.
Otherwise, Sphaeroptica 1.0 will create them for you in a folder placed in the data folder. It will be called "thumbnails".
Sphaeroptica 1.0 will save a JSON project file directly in the data folder. That file will contain all the data needed for Sphaeroptica 1.0 :
- intrinsic matrices
- distortion coefficients
- the extrinsic matrix for each image
- a directory of thumbnails for the virtual camera
The folder structure will look like this :
Command
new command name
Virtual Camera
When you open any project, the first window that you will see is the virtual camera
Screen of the application
On the left, we have the virtual camera, displaying the nearest referenced image (thanks to geometric computations).
On the right, we have :
- Quick buttons to get to a desired view (Frontal, Posterior, Left, Right, Inferior, Superior)
- A list of desired 3D landmarks
- A distance calculator between any two 3D landmarks
How to use the virtual camera ?
With a left click on the image, you will move around a theoretical sphere using geodesic coordinates (Longitude and latitude).
Moving the mouse horizontally and vertically will respectively change the longitude and latitude of the virtual camera.
It is also possible to change those coordinates with the arrows on your keyboard.
Each time the geodesic values of the virtual camera change, Sphaeroptica 1.0 will find the nearest image and display it.
Shortcut views
Sphaeroptica 1.0 allows to have a shortcut to some designated views :
shortcuts
- superior view S
- left view L
- frontal view F
- right view R
- inferior view I
- posterior view P
Left-clicking on any of these widget would move the virtual camera to the desired view.
Right-clicking would set the value of the desired view to the current view and it will automatically save it to the project file.
There is also a keyboard version of these shortcuts :
- Key : same as left-clicking
- Ctrl+Key : same as right-clicking
e.g. pressing F would set the virtual camera to the Frontal view
Landmarks Placements
The main feature of Sphaeroptica 1.0 is the possibility to create landmarks and compute their 3D positions.
For this, we need to configure the landmark and to place it on at least 2 oriented images.
Landmark configuration
You can add as many landmarks as needed for you.
For each landmark it is possible to (in order):
- change its position in the list
- change its label
- change its color
- reset its 3D position and all its references on images
- delete it.
List of landmarks
Place a landmark on an image
If you click on the widget containing the geodesic values, you will display the active image on a new window.
Widget where to click for the popup
window of the landmark-placement page
Choose the landmark you want to place, either by :
- clicking directly on the button with the landmark label
- pressing the + key (goes down the list)
- pressing the - key (goes up the list)
Left clicking on the image will place the designated landmark on the image.
placement of a landmark on an image
When a landmark has been placed on multiple images (minimum 2), Sphaeroptica 1.0 will triangulate its 3D position. That position will then be reprojected on every new image.
reprojection of a 3D landmark
Zoom on the image
You have the possibility to zoom on the image as much as you want to be able to precisely place the landmark at the right pixel.
Here is a list of the actions you are allowed do to zoom in on the image :
- Ctrl++ : zoom 5% in
- Ctrl+- : zoom 5% out
- Ctrl+s : zoom to fit the image entirely in the window
- Ctrl+f : zoom to the 100% resolution of the image
When the image is bigger than the screen, you can scroll on the image with your right click.
Widen the landmarks on the images
It is possible to adjust the size of the landmarks placed on the image. However, be aware that making the landmark bigger, even if it makes it easier to find, makes it harder to know where it is precisely located.
Measurements
When two landmarks have a 3D position, we can measure the distance between them two.
Widgets computing the distances
You can select a defined scale (m, dm, cm or mm, µm, nm), depending on the size of the object.
Additionally, you can rescale all the measurements thanks to a reference. If there is a known distance between two landmarks on the object, you can set it manually by writing it on the distance calculator.
After that, Sphaeroptica 1.0 will automatically modify all the distances computed and it is reversible by pressing the button “Reset Factor”.
Export landmarks into a csv
Clicking the “Export” button would allow you to export all the landmarks in a csv format.
The format will look like this :
Label Color X Y Z X_adjusted Y_adjusted Z_adjusted
with :
- Label, the label of the landmark
- X,Y,Z, the 3D coordinates of the landmarks
- X_adjusted, Y_adjusted, Z_adjusted, the refined coordinates (the coordinates multiplied by the scale factor)
Protocol references
scAnt - Fabian Plum : Link