Image Processing Algorithms Running in Real-time on an Embedded System - Jul 2019
My bachelor thesis research project.
DESCRIPTION
As an Erasmus student I conducted a research project on the performance of computer vision algorithms running on an embedded system, under the supervision of Prof. Antoni Grau Saldes at the VIS - Computer Vision and Intelligent Systems Laboratory.
- Benchmarked feature detectors and descriptors (SIFT, SURF, FAST, ORB, BRIEF) using Python 3 and OpenCV on an Odroid XU4 equipped with a Point Grey industrial camera.
- Ported legacy Python 2.7 OpenCV code to Python 3 and compiled OpenCV from source for the Odroid XU4’s ARM architecture.
- Developed a proof-of-concept system for object detection and tracking with binaural audio feedback, exploring applica-tions in assistive technology for visually impaired users.
This document aims to provide an analysis of the current OpenCV Descriptors and Detectors performance when running in a real-time context.
Understanding the efficiency and effectiveness of these tools is essential for developers and researchers working on computer vision applications.
The performance metrics assessed are speed, accuracy, and reliability in various scenarios.
An early version of the PoC software UI
How the audio feedback worked in this version
I proposed an implementation that would make use of advanced image processing techniques to detect environmental features, such as staircases and other obstacles commonly encountered in daily life, thus enabling users to avoid potential collisions and navigate in safer manner. Furthermore, I suggested an exciting application of these computer vision algorithms in the agricultural sector, where they could be integrated directly into the embedded hardware of drones. This approach would allow for a more efficient and targeted method of pesticide application, ensuring that only the crops requiring treatment would receive the necessary intervention, ultimately promoting sustainability and reducing unnecessary chemical usage while also enhancing crop health.