Baraldi, Andrea (2017) Pre-processing, classification and semantic querying of large-scale Earth observation spaceborne/airborne/terrestrial image databases: Process and product innovations. [Tesi di dottorato]


Download (83MB) | Preview
[error in script] [error in script]
Item Type: Tesi di dottorato
Resource language: English
Title: Pre-processing, classification and semantic querying of large-scale Earth observation spaceborne/airborne/terrestrial image databases: Process and product innovations.
Baraldi, AndreaUNSPECIFIED
Date: 10 April 2017
Number of Pages: 519
Institution: Università degli Studi di Napoli Federico II
Department: Agraria
Dottorato: Scienze agrarie e agroalimentari
Ciclo di dottorato: 29
Coordinatore del Corso di dottorato:
Giannino, FrancescoUNSPECIFIED
Date: 10 April 2017
Number of Pages: 519
Keywords: Earth observation, Big data, Image classification and semantic querying
Settori scientifico-disciplinari del MIUR: Area 01 - Scienze matematiche e informatiche > MAT/08 - Analisi numerica
Date Deposited: 06 May 2017 14:32
Last Modified: 13 Mar 2018 13:25
DOI: 10.6093/UNINA/FEDOA/11724

Collection description

By definition of Wikipedia, “big data is the term adopted for a collection of data sets so large and complex that it becomes difficult to process using on-hand database management tools or traditional data processing applications. The big data challenges typically include capture, curation, storage, search, sharing, transfer, analysis and visualization”. Proposed by the intergovernmental Group on Earth Observations (GEO), the visionary goal of the Global Earth Observation System of Systems (GEOSS) implementation plan for years 2005-2015 is systematic transformation of multisource Earth Observation (EO) “big data” into timely, comprehensive and operational EO value-adding products and services, submitted to the GEO Quality Assurance Framework for Earth Observation (QA4EO) calibration/validation (Cal/Val) requirements. To date the GEOSS mission cannot be considered fulfilled by the remote sensing (RS) community. This is tantamount to saying that past and existing EO image understanding systems (EO-IUSs) have been outpaced by the rate of collection of EO sensory big data, whose quality and quantity are ever-increasing. This true-fact is supported by several observations. For example, no European Space Agency (ESA) EO Level 2 product has ever been systematically generated at the ground segment. By definition, an ESA EO Level 2 product comprises a single-date multi-spectral (MS) image radiometrically calibrated into surface reflectance (SURF) values corrected for geometric, atmospheric, adjacency and topographic effects, stacked with its data-derived scene classification map (SCM), whose thematic legend is general-purpose, user- and application-independent and includes quality layers, such as cloud and cloud-shadow. Since no GEOSS exists to date, present EO content-based image retrieval (CBIR) systems lack EO image understanding capabilities. Hence, no semantic CBIR (SCBIR) system exists to date either, where semantic querying is synonym of semantics-enabled knowledge/information discovery in multi-source big image databases. In set theory, if set A is a strict superset of (or strictly includes) set B, then A  B. This doctoral project moved from the working hypothesis that SCBIR  computer vision (CV), where vision is synonym of scene-from-image reconstruction and understanding  EO image understanding (EO-IU) in operating mode, synonym of GEOSS  ESA EO Level 2 product  human vision. Meaning that necessary not sufficient pre-condition for SCBIR is CV in operating mode, this working hypothesis has two corollaries. First, human visual perception, encompassing well-known visual illusions such as Mach bands illusion, acts as lower bound of CV within the multi-disciplinary domain of cognitive science, i.e., CV is conditioned to include a computational model of human vision. Second, a necessary not sufficient pre-condition for a yet-unfulfilled GEOSS development is systematic generation at the ground segment of ESA EO Level 2 product. Starting from this working hypothesis the overarching goal of this doctoral project was to contribute in research and technical development (R&D) toward filling an analytic and pragmatic information gap from EO big sensory data to EO value-adding information products and services. This R&D objective was conceived to be twofold. First, to develop an original EO-IUS in operating mode, synonym of GEOSS, capable of systematic ESA EO Level 2 product generation from multi-source EO imagery. EO imaging sources vary in terms of: (i) platform, either spaceborne, airborne or terrestrial, (ii) imaging sensor, either: (a) optical, encompassing radiometrically calibrated or uncalibrated images, panchromatic or color images, either true- or false color red-green-blue (RGB), multi-spectral (MS), super-spectral (SS) or hyper-spectral (HS) images, featuring spatial resolution from low (> 1km) to very high (< 1m), or (b) synthetic aperture radar (SAR), specifically, bi-temporal RGB SAR imagery. The second R&D objective was to design and develop a prototypical implementation of an integrated closed-loop EO-IU for semantic querying (EO-IU4SQ) system as a GEOSS proof-of-concept in support of SCBIR. The proposed closed-loop EO-IU4SQ system prototype consists of two subsystems for incremental learning. A primary (dominant, necessary not sufficient) hybrid (combined deductive/top-down/physical model-based and inductive/bottom-up/statistical model-based) feedback EO-IU subsystem in operating mode requires no human-machine interaction to automatically transform in linear time a single-date MS image into an ESA EO Level 2 product as initial condition. A secondary (dependent) hybrid feedback EO Semantic Querying (EO-SQ) subsystem is provided with a graphic user interface (GUI) to streamline human-machine interaction in support of spatiotemporal EO big data analytics and SCBIR operations. EO information products generated as output by the closed-loop EO-IU4SQ system monotonically increase their value-added with closed-loop iterations.


Downloads per month over past year

Actions (login required)

View Item View Item