General Description
External codes and libraries
ANDROMEDE is based on image processing tools available in the OpenCV library: https://docs.opencv.org/4.x/
Some of the numerical methods included in ANDROMEDE are derived from existing image analysis tools. ANDROMEDE provides a graphical interface to these tools. The original scripts of these methods are available on the following sites :
Open and Install
This manual is intended for non-developer users. The opening of the software will be done from the executable provided either in windows version or in LINUX version. The opening of the executable will lead to an independent use of python from the one already installed on the computer. Only useful dependencies for ANDROMEDE are embedded in the executable. The main window can be adapted to the screen size. The top left quadrant is reserved for viewing videos. It will adapt so that the other quadrants remain visible whatever the size of the window or the resolution of the screen (figure description générale des zones de l’interface graphique.).
The opening of a video is done by the menu File >> open (figure ouvrir une vidéo). Several video formats are possible (.avi, .mpg4, .mov ). This command does not load the video in the memory but only creates a video object of openCV. There is therefore no size limitation at this level.
In the following, we will call “frame” the images extracted from the video which are processed one after the other in order to reconstitute the movement of the particles.
Sequence of images
The ANDROMEDE software allows you to process series of images instead of videos. To do this, you must change the file type in the dialog interface. In order to use the same function as for videos, the images will be renamed in ascending order. If you want to keep files with the initial names you must make a copy beforehand. Also, the directory must only contain image files corresponding to the sequence (other types of files are possible). By default, the frequency between images is 1, but it can be modified in the in the dialog box that appears after the image selection.
Processing strategy
The ergonomics of the software is designed to follow the good practices recommended by the developers. The first step is to carry out a complete processing on a reduced number of frames in order to check the relevance of the chosen methods (figure Onglet Paramètres de traitement). The original and processed frames are then loaded into RAM. When the calculation seems relevant, a complete processing of the video can be done. We will then launch the procedure via the menu Process >> Full processing (figure Lancement d’un traitement complet.). The size of the video can be large since the frames are no longer imported into the application. However, it should be noted that the current computing capacities make it possible to process fairly large videos (around one GB) without Full Processing. When the step Full processing is launched, the calculation parameters are those active in the different panels, whether they have already been tested or not.
The document shows the chronological order recommended for the different steps to which a panel or a menu can be associated (in bold in the list below):
Stabilization required if drone, optional otherwise (stabilization)
Choice of the geometric transformation (pre processing)
Test the methods on few images: processing , of detection, of displacement
Save parameters
Processing of the whole sequence with the chosen parameterization (Full processing)
Analysis of the displacements found with possible filtering of outliers and non consistent trajectories (Analysis)
Visualization of the results (Visualization)
Exports of the results in native or GIS format (Export)
Save parameters
The numerical methods can include a significant number of parameters, even if we will see later that the majority of the default parameters are sufficient for an optimal analysis of the videos. It was chosen to keep the parameters accessible because the ANDROMEDE tool is designed to adapt to all possible image-taking configurations and study objectives. Thus, for cases other than surface velocities, modifying the parameters will allow to reach the objectives. Other studies already using this type of algorithm include fish tracking, granular media studies (avalanches), ice or lava advance, terrestrial fauna displacement (insects), hydrodynamics in the laboratory.
The parameters are stored in dictionaries corresponding to the steps described in the previous paragraph. It is possible to save the parameters in text format using the command File>>Write/Export>>Parameter. The parameters can then be reloaded automatically with the command File>>Read/Import>>Parameter (figure Paramètres d’import/export). The recording allows to do identical treatments on several videos but also to keep a trace of the calculations made. The parameters file also allows you to modify values that may not be yet available in the interface.
Interface using
When the video is loaded, it is possible to know the information of the displayed frame by double clicking on a pixel. In the dialog box at the bottom right, the position of the pixel in X (horizontal direction) and in Y (vertical direction) appears, followed by the values of the color bands (RGB usually). It is possible to zoom with the mouse wheel. The center of the zoom area is the position of the cursor on the image. When you double-click on the zoomed image, the information that appears corresponds to the position in the initial image (same for processed images).
The parameters are automatically saved as soon as they are modified.
It is possible to display each frame of the video by moving with the progress bar under the graphic window or with the buttons Previous and Next. A scrolling menu under the image allows you to navigate between the different steps of the calculation and the associated images. When loading the video, only the option original is possible.