engleski

An assessment of social media oil spill reports using transfer learning on remotely sensed images

Nowadays, we are increasingly aware that irrespon- sible human behavior is the main reason for many instances ofenvironmental pollution, including oil spills in the sea. In orderto detect such contaminants in a timely manner and take care ofthem as quickly as possible, in this paper we present a methodfor automatic oil spill detection through remote sensing overSentinel-1 SAR images. Presented approach uses social mediahuman reports of oil spill as input information for building adata set and training a model over collected images of actualoil spills. Because of a rather small oil spill data set, we areusing two methods for obtaining better results. Namely, we usedata set augmentation for enrichment of images in the data setand transfer learning to retrain our collection of images basedon trained deep network on ImageNet data set. For classifyingimages we used four machine learning classification modelswith different accuracy in oil spill detection. A comparison ofaccuracies of the individual machine learning model and thewhole process of oil spill detection initiated by social mediareports and further validated by constructed classifier will bedescribed in this paper.

Oil spill ; Twitter ; Deep learning ; Transfer learn-ing ; SAR ; Sentinel-1

nije evidentirano