Daniele De Vecchi

Automatic clouds/shadows extraction method from CBERS-2 CCD and LANDSAT data

Satellite acquisitions from LANDSAT (LS) and CBERS programs are widely used in monitoring land cover dynamics. In the acquired products, clouds form opaque objects are obscuring parts of the scene and preventing a reliable extraction of information from these areas. Consequently, cloud shadows create similar problems, as the reflected intensity of the shadowed areas is highly reduced, generating additional info gaps. The problem can be handled by replacing clouds/shadows pixels from other close-date acquisitions, but that would assume a prior knowledge of the spatial distribution of clouds and their corresponding shadows in a scene. This research introduces a method that provides the clouds/shadows layers and their percentage in LS (TM & ETM+) and CBERS (HRCC) scenes. The approach relies on a set of literature indicators to create a composite image that enhances the visual differentiation of clouds/shadows from other objects. The created composite RGB are then warped to a relative luminance raster calculated from the linear bands components. Afterwards, the raster is processed by a K-means unsupervised classifier with a definite number of classes in order to isolate the target-layer pixels. Next, the statistical mode for the population of each class is calculated, compared and used to select the cloud/shadow class automatically, and finally the results are refined by a set of morphological filters. The processing chain avoids the usage of thresholds and highly reduces the user intervention. The achieved outcomes on various test cases are promising and stable, and encourage further developments.

CLOOPSy - Copernicus Land cOver crOwdsourcing Platform for Sentinel-based mapping

CLOOPSy is a mobile application designed with the aim to collect data from a crowd of volunteers (Dell’Acqua and De Vecchi 2017). In particular, the idea is to ask them to contribute with a picture (Burgelman 2015)(Marr 2016) and to select a land cover class based on the CORINE taxonomy (Laso Bayas et al. 2016). Collected reports, following the administrator approval, will be released to the public and available for download using the APIs. A tutorial is mandatory for each new user. Reports will be used to automatically validate of algorithms working on satellite data, update of land cover layers and to train machine learning algorithms. CLOOPSy is a native mobile app developed using the Xamarin framework (Xamarin 2017). Reports can only be submitted by registered users. Collected pictures are seasoned with GNSS location and compass direction. Examples for each class of the CORINE Land Cover taxonomy are provided with the aim to help in the decision. The app is built on a general framework, therefore it can be adapted in order to collect different things. Every submitted report is sent to a remote server. It is available for Android on the Google Play store and soon it will be released on the Apple App store. A web portal is also available for CLOOPSy. Volunteers can register and begin to submit reports using the ‘Register’ page. The main page is a map showing all the public reports and it does not require any login. Reports are considered public only after being approved by super-users. Once a user is logged, the map is filled with all the reports, including those submitted by him/herself and not yet reviewed. The server integrates an algorithm for automatic matching with parcel GIS layers where available. The service is hosted by the ESA RSS (Research and Service Support) service (Agency 2017) for future easy integration with Sentinel-2 data repository. Public and private reports can be download using RESTful (Representational state transfer) APIs.

A feature-based approach to register CBERS CCD and HRC imagery for built-up area extraction purposes

The open data policy, the availability of high resolution imagery and the capability to cover fast-growing economies are among the main advantages of CBERS. Unfortunately, data produced by this satellite suffer of geographic misplacement, forcing to apply pre-processing techniques to stabilize the imagery. This paper introduces a feature-based technique developed to pre-process CBERS imagery over an area of interest. In particular, the algorithm is able to fix the shift among HRC high resolution (2.5 meters) and CCD medium resolution (20 meters). The final goal is to combine the advantages of high resolution and radiometric properties for built-up area extraction purposes.

A PCA-based hybrid approach for built-up area extraction from Landsat 5, 7 and 8 datasets

Urban expansion monitoring and organization can be performed through space-based observation thanks to the revisit time and level of details guaranteed by satellite remote sensing. In particular, the Landsat mission products are the most used thanks to the long time coverage and open access policy. This paper proposed a hybrid method - combination of pixel- and object-based analysis - in order to automatically extract built-up areas from Landsat imagery. Segments are delineated from spectral indices computed in order to increase the spectral distance among the different land cover classes. The principal component analysis is applied to the original bands and constitutes the pixel-based side of the method. Segments and PCA are then combined and classified using an unsupervised approach. Results of the method were quite satisfying with an average Kappa value over 0.5 in both case studies.

A small step towards the citizen sensor: A multi-purpose framework for mobile apps

The concept of crowdsourcing is to collect and share information from “the crowd”. The diffusion of smartphones and tablets led to what can be pictured as a “dense network of observers” that can be used to submit useful data possibly complementing satellite-based Earth observation. In this paper, we propose a multi-purpose framework specifically designed to collect data in a distributed way; the model is based on a client-server architecture and an open-source framework including PHP and SQL. Communication among the devices is guaranteed by standard protocols like HTTP and JSON. An example of application based on this framework is also presented; the main objective is to collect information about water quality by asking users to provide feedback. The interface is both web- and mobile-based.

Potentials of Active and Passive Geospatial Crowdsourcing in Complementing Sentinel Data and Supporting Copernicus Service Portfolio

The recent trend toward open Earth observation (EO) data has revived a general interest in satellite-based monitoring and mapping of the Earth surface. The open policy now applied to LANDSAT data, and the starting of Sentinel operations, whose data are freely distributed even for commercial purposes, tore down a financial barrier to wider use of EO data in business activities, especially those with a narrow financial margin. Notwithstanding the flood of open data, some aspects of the Earth surface still escape satisfactory monitoring from space, especially in complex, anthropic areas. Due to insufficient spatial resolution, lack of visibility, or unsuitable revisit times, important pieces of information may not emerge from spaceborne data. In situ sensing can represent a vital source of integrative information to fill the aforementioned gaps and build a more complete and accurate picture of the situation and trends in the observed area. The contribution of in situ sensing was envisaged quite early in the Earth observation history, but for a long time it remained limited to tailored sensors displaced in strategic locations. With the increasing circulation of smartphones, a new opportunity has recently opened for a different paradigm of in situ sensing, offering a huge mass of additional data by tapping on data generated by mobile devices. Even if such data may be less specialized and less usable for various reasons, the sheer size of the data flow ensures that statistical analysis will pick possible useful clues. The increasing availability of mobile connections has indeed revived the concept of “crowdsourcing,” i.e., entrusting a pool of actors with problem solution or information collection tasks. In our scenario, individuals carrying mobile devices can become “citizen sensors” on a voluntary basis by contributing data through their connected terminals. Even considering the typical issues of crowdsourced data, like quality and reliability, the balance remains definitely positive. This paper provides an overview of the theme and discusses how it relates to an important, coordinated EO initiative like Copernicus. It finally presents a specific example realized in the framework of a recent research project under the Copernicus aegis.

Unsupervised change detection for urban expansion monitoring: An object-based approach

Change detection is by definition the capability to detect and highlight changes occurring in space and time. Earth Observation satellites represent a fundamental source of information thanks to repeatability in time and spatial resolution. In this paper, we propose an unsupervised change detection technique capable of processing a series of single-date built-up area extractions with two main goals: determining the age of different parts of an urban area and fixing errors due to the automatic extractions suggested in previous papers by our group. Results show a general stabilization of the Kappa value but further investigation is still necessary. The proposed algorithm is available to the general public as a part of a QGIS plugin named SENSUM Earth Observation (EO) tools.