Abstract
Automatic supervised classification with complex modelling such as deep neural networks requires the availability of representative training data sets. While there exists a plethora of data sets that can be used for this purpose, they are usually very heterogeneous and not interoperable. In this context, the present work has a twofold objective: (i) to describe procedures of open-source training data management, integration, and data retrieval, and (ii) to demonstrate the practical use of varying source training data for remote sensing image classification. For the former, we propose SatImNet, a collection of open training data, structured and harmonized according to specific rules. For the latter, two modelling approaches based on convolutional neural networks have been designed and configured to deal with satellite image classification and segmentation.
Highlights
Data-driven modelling requires sufficient and representative samples that capture and convey significant information about the phenomenon under study
We propose a methodology to organise open and freely available training data sets designed for satellite image classification in view of fusing them with other Earth observation (EO)-based products
Within SatImNet context, we investigate data fusion by conducting experiments demonstrating a satellite image classification and segmentation application based on CNN models
Summary
Data-driven modelling requires sufficient and representative samples that capture and convey significant information about the phenomenon under study. For instance, there exist large collections of pre-trained models dealing with image classification [5,6]. The collection of good quality training sets for supervised learning is an expensive, error-prone [7], and time-consuming procedure It involves manual or semi-automatic label annotation, verification, and deployment of a suitable sampling strategy like systematic, stratified, reservoir, cluster, snowball, time-location, and many other sampling techniques [8]. The joint use of two or more existing training data sets require careful examination of their individual features and organisation To ease this process, we propose a methodology to organise open and freely available training data sets designed for satellite image classification in view of fusing them with other Earth observation (EO)-based products.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have