Aerial imagery analysis, processing, and presentation scripts.
I love the University of Minnesota, but my funding ran out and we weren't able to get anything new lined up in time. So March 18 is my last day at the University of Minnesota, Dept. of Aerospace Engineering, UAV lab. :-(
At the same time I'm super excited to be transition to a new position at Cirrus Aircraft in Duluth. This will change many things, including my time available and my focus relative to working on this project. For a couple years this code was integral to my daily work, and that probably will no longer be true once I transition to my new position.
My hope is to keep this project going long term, to be available (within reason) to answer questions, and to guide people getting up to speed for the first time. I am sincerly hopeful I can continue to improve and add features as time permits.
A mosaic of 2812 images from a Phantom 4 Pro v2 mission
A 3d surface rendering of the same area
A zoom in showing all the original detail and oblique perspective from the
original imagery because the mapping tool simply draws the original images in a
big pile all sized, oriented, stretched, and aligned perfectly.
This image analysis tool-set started as two independent efforts to build a better (faster, free-er, more robust) image stitching system for aerial surveying. The accumulated knowledge of these two projects were merged to form a single project going forward. Over time, this project has evolved into a full blown aerial survey and mapping system that supports a variety of research projects with specific needs (use-cases) not well supported by existing commercially tool. The ImageAnalysis project improves over most existing systems in several distinct ways:
-
The final orthographic map as presented as a big pile of original images; all sized, scaled, rotated, stretched, and fitted to each other perfectly. All the original images, perspectives, and full pixel resolution are preserved and can be examind in the context of the full map. For our projects, this is a powerful tool for "finding a needle in a haystack" type tasks.
-
New robust match finding strategies have been developed that find a significantly higher number of image match pairs in traditionally challenging environments such as forests or mature crop fields. We regularly out-perform well known commercial packages in terms of fitting more of your images together correctly into the final map.
-
All the code is licensed with the MIT open-source license and written in python to be as open and accessible as possible. We are developing this project in a research lab and wish to be open and share the knowledge and tools we develop.
The project goals remain:
-
develop a quality open-source image fitting (stitching) and scene reconstruction tool chain that is appropriate for examination (education) and modification (research use.)
-
leverage well known building blocks including python, opencv, numpy, and scipy.
-
continue to develop and refine offshoot projects that support our UAS Lab projects: such as EKF validation, augmented reality, movie/flight-data time correlation.
-
develop extensions and improvements in support of ongoing research project. (Such as: invasive species surveys, mapping, and ground truthing.)
This is likely to be out of date faster than I can write it, but here are a few things I have been working recently.
-
Improving and streamlining the project structure layout. Now an "ImageAnalysis" folder is added to each folder of raw images.
-
Improving the project setup flow and autodetecting camera/lens config and image geotags when possible.
-
Streamlining support for DJI and Sentera camera systems.
-
Continuing visual presentation improvements
-
Signficant improvements finding matches in challenging areas (that break many commercial tools), areas such as forests and mature crop fields when imaged from a low altitude.
-
Continuing work on image connection grouping optimizations to improve solver (optimizer) results and reduce artifacts.
-
[done] Renable/retool code to best fit optimized solution to original camera pose locations. (This ensures the solution didn't walk off in some random direction, and also that the final fit is as close to the real world location as we are able to do with the given data set.)
-
[done] Continued improvement of python3 support
-
[done] A scene exploration tool that knows all the views covering any point, can geolocate any selected point in an image, can import/export kml/csv for sharing results or to export to other mapping tools.
Briefly, near term development goals include:
- Continue work on real-time shader language techniques to highlight different types of (color) features in the imagery.
Medium term development goals include:
- Continued exploration of machine vision strategies for automatically identifying/locating items of interest in imagery. (Such as: oriental bittersweet.)
Wish list items:
-
The current code optimizes the 'sparse' mesh, but does not do dense mesh reconstruction. It would be interesting to explore dense mesh generation for creating detailed DEM's and exact orthophotos. This is a core strength of commercial image stitching tools, but perhaps not necessary for every aerial survey application.
The most important objective is to answer questions like "Where is all the oriental bittersweet in the survey area?" A pretty picture, an orthorphoto, or a DEM may not always be the best tool to help answer that question, while other customized approaches may do better.
-
I would love to see a community of interested tinkerers grow up around this project. I think there are many important concepts and strategies here that have far ranging uses and implications. It would be great to keep a light shining on this area so the expertise doesn't get entirely buried inside the caves of the commercial world.
Home to 3rd party code that needs modifications or adjustments to be helpful. (Or things that aren't commonly available system-wide.)
A series of front-end scripts that primarily pair with the lib directory for feature detection, matching, and scene assembly. For making maps the process is actually quite simple:
-
run "process.py /path/to/images" This will process all the images and create the map.
-
run "explorer.py /path/to/images" This will launch the interactive explorer tool to view your map.
Libraries of python modules with all the functionality that supports the higher level scripts.
For the purposes of generating an initial earth surface estimate, use SRTM data. Given a camera pose estimate (i.e. from flight data) and an earth surface estimate, we can project out the feature vectors and estimate their 3d locations. This is useful for generating an initial guess to feed the sparse optimizer (sparse bundle adjustment.)
A random collection of scripts for testing different things.
- ils: This relates to systems that have an illumination sensor pointing up at the sky. When the drone pitches or rolls for turning or forward motion, the sensor no longer points up. This code understands date, time, location, as well as the relative sun location and angle. It attempts to correct the illumination sensor for attitude errors and thus produce more consistent results in the output images.
Some of these image analysis techniques can be applied to movies in interesting ways.
-
Use feature matching between consecutive movie frames to accurately estimate a gyro axis (aligned with the camera center of projection.) Will also estimate the 2nd and 3rd gyro axes, but with less quality.
-
Track Aruco codes. If you have control over your scene and can place Aruco codes in strategic places, they are extremely awesome. The detection code is extremely fast, very reliable, each marker has it's own code, and all 4 corners of the marker are identified.
-
Extract still shots (frames) from a movie and geotag them.
-
Generate an augmented reality hud overlay on top of an in-flight movie.
- pip install opencv-contrib-python (still needed?)
- pip install panda3d