The overall aim underlying this project is the development of RPAS based beyond visual line of sight (BVLOS) surveillance infrastructure and technology, including automated data analysis using Artificial Intelligence (AI). The envisioned infrastructure and technology can be used for a variety of purposes such as the surveillance of highways and other infrastructure, slope stability analysis, search and rescue and wildlife detection and biodiversity monitoring.
We have developed and trained convolutional neural networks (CNN, a form of AI) to automate the recognition of debris on images. This required the collection and labelling for the presence / absence of debris of large amounts of data and the application of data augmentation to artificially increase the number of images used for training. For video material we focused on applying a state-of-the-art detector (YOLO V3) to detect amongst others traffic signs, vehicles, buildings and people.