I found a document detailing how to code a drone using image recognition. I decided to leave a note for a closer look when the need arises.
The source code is also available on github.
Master's thesis http://digitalassets.lib.berkeley.edu/techreports/ucb/text/EECS-2014-117.pdf
Github https://github.com/ssk2/drones-267
If you look at the Github page, papers presentations You can read and recommend more. image segmentation
Bryan Catanzaro et al. [Drones-267 / papers / Efficient, High-Quality Image Contour Detection.pdf](https://github.com/ssk2/drones-267/blob/f6920127ef8c9a5ef0d20719a7bb1fca50bad46d/papers/Efficient%2CHigh-Quality%20Image% 20Contour% 20Detection.pdf) To reach the dissertation.
Since it is a field that requires real-time performance, the idea of SIMD and multithreading is mentioned. As of 2014, it describes what kind of ARM-based board was selected, what kind of library was replaced and used, and how much difference in processing time was caused by the selection of method. ..
24 corner points. It solves 48 equations with 6 degrees of freedom by SVD (Singular Value Decomposition). He wrote that the detection of the corner point was hotpoint.
BeagleBone Black Hardkernel Odroid XU Robotics Operating System (ROS), Etc. are used.
Estimate the camera position using the AR marker. Github Aruco-with-Python https://github.com/JanezCim/Aruco-with-Python
Apparently I made a Python binding to the Aruco library.
https://www.arnavichara.com/ar_about2.html Commentary article on AR markers
Investigating ArUco, a lightweight AR library
There is also a trend toward open source for drones
Recommended Posts