The integrated use of polarimetric Synthetic Aperture Radar (SAR) and optical image data for land cover mapping using an object-based approach

De Beyer Leigh Helen. 2015. The integrated use of polarimetric Synthetic Aperture Radar (SAR) and optical image data for land cover mapping using an object-based approach. Stellenbosch : Stellenbosch University, 184 p. Thesis MSc : Science in Geoinformatics : Stellenbosch University

Encadrement : Kemp, Jaco ; Todoroff, Pierre

Résumé : Image classification has long been used in earth observation and is driven by the need for accurate maps to develop conceptual and predictive models of Earth system processes. Synthetic aperture radar (SAR) imagery is used ever more frequently in land cover classification due to its complementary nature with optical data. There is therefore a growing need for reliable, accurate methods for using SAR and optical data together in land use and land cover classifications. However, combining data sets inevitably increases data dimensionality and these large, complex data sets are difficult to handle. It is therefore important to assess the benefits and limitations of using multi-temporal, dual-sensor data for applications such as land cover classification. This thesis undertakes this assessment through four main experiments based on combined RADARSAT-2 and SPOT-5 imagery of the southern part of Reunion Island. In Experiment 1, the use of feature selection for dimensionality reduction was considered. The rankings of important features for both single-sensor and dual-sensor data were assessed for four dates spanning a 6-month period, which coincided with both the wet and dry season. The mean textural features produced from the optical bands were consistently ranked highly across all dates. In the two later dates (29 May and 9 August 2014), the SAR features were more prevalent, showing that SAR and optical data have complementary natures. SAR data can be used to separate classes when optical imagery is insufficient. Experiment 2 compared the accuracy of six supervised and machine learning classification algorithms to determine which performed best with this complex data set. The Random Forest classification algorithm produced the highest accuracies and was therefore used in Experiments 3 and 4. Experiment 3 assessed the benefits of using combined SAR-optical imagery over single-sensor imagery for land cover classifications on four separate dates. The fused imagery produced consistently higher overall accuracies. The 29 May 2014 fused data produced the best accuracy of 69.8%. The fused classifications had more consistent results over the four dates than the single-sensor imagery, which suffered lower accuracies, especially for imagery acquired later in the season. In Experiment 4, the use of multi-temporal, dual-sensor data for classification was evaluated. Feature selection was used to reduce the data set from 638 potential training features to 50, which produced the best accuracy of 74.1% in comparison to 71.9% using all of the features. This result validated the use of multi-temporal data over single-date data for land cover classifications. It also validated the use of feature selection to successfully inform data reduction without compromising the accuracy of the final product. Multi-temporal and dual-sensor data shows potential for mapping land cover in a tropical, mountainous region that would otherwise be challenging to map using single-sensor data. However, accuracies generally remained lower than would allow for transferability and replication of the current methodology. Classification algorithm optimisation, supervised segmentation and improved training data should be considered to improve these results.

Résumé (autre langue) : Beeld-klassifikasie word al 'n geruime tyd in aardwaarneming gebruik en word gedryf deur die behoefte aan akkurate kaarte om konseptuele en voorspellende modelle van aard-stelsel prosesse te ontwikkel. Sintetiese apertuur radar (SAR) beelde word ook meer dikwels in landdekking klassifikasie gebruik as gevolg van die aanvullende waarde daarvan met optiese data. Daar is dus 'n groeiende behoefte aan betroubare, akkurate metodes vir die gesamentlike gebruik van SAR en optiese data in landdekking klassifikasies. Die kombinasie van datastelle bring egter 'n onvermydelike verhoging in data dimensionaliteit mee, en hierdie groot, komplekse datastelle is moeilik om te hanteer. Dus is dit belangrik om die voordele en beperkings van die gebruik van multi-temporale, dubbel-sensor data vir toepassings soos landdekking-klassifikasie te evalueer. Die waarde van gekombineerde (versmelte) RADARSAT-2 en SPOT-5 beelde word in hierdie tesis deur middel van vier eksperimente geevalueer. In Eksperiment 1 is die gebruik van kenmerk seleksie vir dimensionaliteit-vermindering toegepas. Die ranglys van belangrike kenmerke vir beide enkel-sensor en 'n dubbel-sensor data is beoordeel vir vier datums wat oor 'n tydperk van 6 maande strek. Die gemiddelde tekstuur kenmerke uit die optiese lae is konsekwent hoog oor alle datums geplaas. In die twee later datums (29 Mei en 9 Augustus 2014) was die SAR kenmerke meer algemeen, wat dui op die aanvullende aard van SAR en optiese data. SAR data dus gebruik kan word om klasse te onderskei wanneer optiese beelde onvoldoende daarvoor is. Eksperiment 2 het die akkuraatheid van ses gerigte en masjien-leer klassifikasie algoritmes vergelyk om te bepaal watter die beste met hierdie komplekse datastel presteer. Die random gorest klassifikasie algoritme het die hoogste akkuraatheid bereik en is dus in Eksperimente 3 en 4 gebruik. Eksperiment 3 het die voordele van gekombineerde SAR-optiese beelde oor enkel-sensor beelde vir landdekking klassifikasies op vier afsonderlike datums beoordeel. Die versmelte beelde het konsekwent hoër algehele akkuraathede as enkel-sensor beelde gelewer. Die 29 Mei 2014 data het die hoogste akkuraatheid van 69,8% bereik. Die versmelte klassifikasies het ook meer konsekwente resultate oor die vier datums gelewer en die enkel-sensor beelde het tot laer akkuraathede gelei, veral vir die later datums. In Eksperiment 4 is die gebruik van multi-temporale, dubbel-sensor data vir klassifikasie ge-evalueer. Kenmerkseleksie is gebruik om die data stel van 638 potensiële kenmerke na 50 te verminder, wat die beste akkuraatheid van 74,1% gelewer het. Hierdie resultaat bevestig die belangrikheid van multi-temporale data vir grond dekking klassifikasies. Dit bekragtig ook die gebruik van kenmerkseleksie om data vermindering suksesvol te rig sonder om die akkuraatheid van die finale produk te belemmer. Multi-temporale en dubbel-sensor data toon potensiaal vir die kartering van landdekking in 'n tropiese, bergagtige streek wat andersins uitdagend sou wees om te karteer met behulp van enkel-sensor data. Oor die algemeen het akkuraathede egter te laag gebly om vir oordraagbaarheid en herhaling van die huidige metode toe te laat. Klassifikasie algoritme optimalisering, gerigte segmentering en verbeterde opleiding data moet oorweeg word om hierdie resultate te verbeter.

Mots-clés libres : Land cover classification, RADARSAT-2, SPOT-5, Object-based, Feature selection

Auteurs et affiliations

• De Beyer Leigh Helen, Stellenbosch University (ZAF)

Source : Cirad-Agritrop (https://agritrop.cirad.fr/589971/)

Métriques

[ Page générée et mise en cache le 2023-08-29 ]