pennyhow/pytrx

PyTrx is a Python object-oriented programme created for the purpose of calculating real-world measurements from oblique images and time-lapse image series. Its primary purpose is to obtain velocities, surface areas, and distances from oblique, optical imagery of glacial environments.

pytrx's Issues

Better documentation and commenting in scripts

DEM.py needs looking at mostly, but also the other scripts could use a little more commentary too.

Check TestPlotBands.py and TestPlotCalibrated.py

These scripts could be used as an example driver for PyTrx.

.CR2 raw image parsing

Many requests involve analysing RAW images. Currently, PyTrx is compatiable with JPG and PNG images. This is largely because of differences in the image EXIF information, which PyTrx uses to gather information about each image. Routines are needed to ensure that RAW images can be ingested with compatible information.

Bug in defineColourRange when used in Area object

Modify so that a given colour range is assigned to the Area object as Area._colourrange.

Migrate functions from CamEnv_old.py to CamEnv.py

Some of the functions from the old script have not yet been moved to the new script.
Specifically:

1. getName
2. getCalibration
3. getCameraPose
4. getFullCamera (?)
5. showGCPs
6. showPrincipalPoint
7. approxYPR
8. Yaw
9. PitchRoll
10. voxelviewshed (?)

Also, migrate any useful functionality from the Camera class which is currently commented out in the CamEnv.py script (above the CamEnv class).

CameraEnvironmentData_KR2_2014.txt Calibration Question

Hi Penny, what do the 6 and 9 refer to in line 24 of CameraEnvironmentData_KR2_2014.txt when setting the calibration image path?

[../Examples/camenv_data/calib/KR2_calibimgs/*.JPG, 6, 9]

driver_velocity2-Calculate velocities

hello sir,In the example driver_velocity2 ，the Calculate velocities, I found that the method parameter is not available. I want to know why ？

Cannot run calcDenseVelocity with campars=None

PyTrx/PyTrx/Velocity.py

Line 674 in cd614be

xyz0, uv0 = seedGrid(campars[0], griddistance, campars[1], mask)

calcDenseVelocity should be able to run with campars = None but, campars is indexed as a list therefore runs into an error if it is not supplied. I wanted to run calcDenseVelocity on two orthoimages resampled in the same grid, and therefore did not need any camera parameters or DEMs. The docstring suggests that the function can be run without campars. It should be stated clearly whether or not campars is needed for the function.

Merge apply_persp_homography functions into one function

Marked in Images.py script, in the TimeLapse class.

Modules importing incorrectly for readthedocs

OpenCV is having import issues with readthedocs. PyTrx modules that import cv2 bring up this error in the readthedocs compiling:

AttributeError:` partially initialized module 'cv2' has no attribute 'gapi_wip_gst_GStreamerPipeline' (most likely due to a circular import)

Long-term PyTrx to-do list

• Better manual delineation functionality in the Area and Line objects (e.g. eliminate the risk of infinite looping)
• Faster georectification processing. This can either be done by sourcing more efficient methods, or potentially using parrallel processing
• Faster handling of DEM, so that users that input DEMs with higher spatial resolution and coverage

Migrate functions from Images_old.py to Images.py

Namely the checkMask and getAllInfoNo functions. Check to see if there are others that need to be migrated.

Better constraining and documentation of georectification viewshed flag

Resolve viewshed limitations within the CamEnv.setProjection function. The viewshed is computed from the camera position and the DEM to ensure that image points are only projected to visible parts of the DEM surface. However, sometimes this is not needed - either the DEM has errors or the reprojection is highly sensitive to elevation changes.

An alternative would be to try and compute the projection with the viewshed, but resort to disregarding it if this does not compute. And also, this should be better documented within the readthedocs (maybe within an example driver script)

TimeLapse functionality for processing through an entire sequence, not just one image

The Image.TimeLapse class currently only has functions for processing one image pair. Functions should also be available for processing through an entire image sequence. This has been done in other scripts, but it would be better if these were in the TimeLapse class.

Develop CamEnv class so that it can be initialised with raw input

Develop CamEnv class so that it can be initialised with raw input, thus it doesn't entirely rely on a .txt file containing the camera environment data directories.

Consult Aslak and Alex about the transformation functionality

There are a couple of comments in the CamEnv class where there has been a translation from ImGRAFT, and one specific instance where it is questioned why a certain step has been used. These are specifically in the projection and inverse projection functions.

'ValueError: No Points Given' when running projectUV

In my forked repo, in line 53 running projectUV returns a ValueError: No Points Given. Tracing this to the setProjection function in PyTrx/CamEnv.py, lines 852-862, before the projection, X, Y, and Z are all arrays with points, after the projection they are empty arrays. Thank you for any help you can offer.

Calculate camera model from focal length, image size and sensor dimensions

Estimate camera model from a given focal length (mm), image size (px) and physical sensor dimenstions (mm); as described here

Functions for calculating volume loss from sequential terminus positions

Implement functions used for Harold's bachelors dissertation.

** I have asked Harold for a copy of his Python code.

Test functionality for colored_bar and plotTrackRose in Utilities.py

Currently it looks like these functions are not really used, and do not use real data inputs.

Extraction of spot velocities/spatial averages from velocity fields

And the ability to write georeferenced velocity fields to tif files for use in ArcGIS and QGIS.

Explore other interpolation methods

Related to issue #1
Explore other, more robust ways to interpolate and make velocity fields. Perhaps something that weights the interpolation based on point density?

Functions for extracting velocities to an output file

Currently do not exist. Possibly write this functionality into the FileHandler module.

Check checkMatrix function to see if Matrix input is compatible

Marked in script (in CamEnv.py).

Integrate functions from MadlebrotText.py

For better visualisation of subplot data.

Implement interpolation functions

Interpolation functions for making velocity fields currently are absent. They should be part of the Utilities and TimeLapse scripts. It is intended to first transpose these functions from Lynne's version of PyTrx.

PyTrx not in conda environment dependencies?

Hi Penny,

I was wondering why PyTrx is not included in your environment.yml?

Thank you a lot in advance and have a nice day!

Implement Measure.py module into the rest of PyTrx

Currently the Measure module works off of functions and classes from an older version of PyTrx. The Area and Length classes can also possibly be refined, as they are a little long-winded at the moment.

Different methods to set timings in the TimeLapse object

The set_timings function in the TimeLapse object should be able to be set using different methods other than from the EXIF info.
Proposed method: timings defined from specified time increments.

Change CamCalib initialisation so no data can be inputted and carried forward

Change CamCalib object so that 'None' can be passed, which will initialise an empty CamCalib object containing no calibration data (this could be useful for if we wish to use PyTrx for orthphotos and/or satellite imagery).

Better visualisation of points on oblique images, points on DEMs and velocity fields

This is all currently in Utilities.py. The subplots need better presentation, and plotting functions for velocity fields are also needed (there are some in Lynne's version of PyTrx.Velocity).

Compile FileHandler.readDEM functions into one master function

The functions readDEM, readDEMxyz and readDEMmat could potentially be compiled into one function that automatically detects the file type and imports the data based on the file type.

Use some of the functions in Development.py to form example drivers

E.g. Homography calculations, velocity calculations, plotting.

Change YPR guidance

Currently, we advise users to use camera yaw, pitch, roll pose (YPR) values of

[0, 0, 0] #[yaw, pitch, roll]

if the parameters are unknown. This can either be set in the camera environment txt file, or as a variable in a driver script. However, this causes problems in the optimisation routine - if the yaw parameter is 0 then the optimisation routine may not be able to resolve the camera environment from the ground control points. It should be advised to use an estimate of the yaw parameter in order to give the optimisation routine a starting point e.g.

[90, 0, 0] #i.e. camera pointing east

Finish or discard apply_cam_correction_points function (in Images.TimeLapse)

It is currently just the first line of the function, which is commented out (marked in script).
Don't know if it is needed or whether it can be discarded.

Add more checks to the Image.TimeLapse reporting function

I have marked this in the Images.py script.

Contain more initialisation options for Image.ImageSequence

It's commented in the script that there should be an option to initialise an ImageSequence object from a file which contains the names of the images.