Ocenjevanje vlažnosti iz radarskih slik SAR

doktorska disertacija

Matej Kseneman (Author), Dušan Gleich (Mentor), Žarko Čučej (Co-mentor)

Abstract

Doktorska disertacija podaja ocenjevanje vlažnosti tal iz podatkov radarja z umetno odprtino (ang. Synthetic Aperture Radar - SAR), kar se izvaja s pretvorbo koeficienta povratnega sipanja radarskih mikrovalov v relativno dielektrično konstanto, preko katere se le-ta pretvori v nam razumljivejšo pro-storninsko vlažnost tal. Novost, ki jo podajam v doktorski disertaciji, jevpeljava novega modela na področju ocenjevanja vlažnosti tal. Predlagani model deluje v pasu X radarskih mikrovalov, pri katerem smo poleg odprtih površin upoštevali tudi prekritost z nizko vegetacijo. Delovanje algoritma ocenjevanja vlažnosti tal primerjamo z merjenimi podatki terenskih meritev, medtem ko koeficiente povratnega sipanja pridobimo iz visoko-ločljivih TerraSAR-X slik SAR z različnimi parametri radarja SAR, kar zajema različne vpadne kote, polarizacijo in razdaljno pasovno širino. V sklopu disertacije predlagamo dva modela, izmed katerih je prvi osnovan na Shi pol-empiričnemu modelu in deluje le na odprtih površinah oz. področjih brez prisotnosti vegetacije. Ta model modifici-ramo z vpeljavo novih koeficientov modela, ki jih določimo z regresijskimi metodami v kombinaciji s terenskimi meritvami, kjer uporabimo algoritem minimalne povprečne kvadratne napake. Drugi, splošnejši model, ki deluje tudi na področjih prekritih z nizko vegetacijo, jeosnovan na modelu voda-oblak , kateremu je v osnovi dodan še dodaten t.i. vegetacijski člen, ki nam omogoča ocenjevanje vlažnosti tal tudi na predelih, kjer se nahaja nizka rast vegetacije. Ocenjevanje vlažnosti je poleg parametrov radarja odvisno še od hrapavosti površja, le to pa smo upoštevali vobliki parametra RMS višine. Preostala parametra predstavljata še delno prekritost z vegetacijo in Fresnelov refleksijski koeficient. Ker ti parametrivnaprej niso znani, smo model zapisali v matrični obliki, tako da lahko model vpeljemo v iterativno shemo regularizacije Tikhonova. Algoritem zaocenjevanje neznanih parametrov tal izhaja iz dobrih lastnosti regularizacije slik, kjer želimo s parametri regularizacije odpraviti ali kompenzirati efekte korelacijske dolžine in korelacijske funkcije. Eksperimentalni rezultati so pokazali uporabnost obeh modelov, vendar primerjava modelov kaže, da pri rezultatih prednjači predlagan dvokomponentni model, osnovan na modelu voda-oblak z iterativno regularizacijo Tikhonova.

Keywords

ocenjevanje vlažnosti tal;TerraSAR-X satelit;regularizacija Tikhonova;empirični model;

Data

Language:	Slovenian
Year of publishing:	2012
Source:	Maribor
Typology:	2.08 - Doctoral Dissertation
Organization:	UM FERI - Faculty of Electrical Engineering and Computer Science
Publisher:	[M. Kseneman]
UDC:	004
COBISS:	262816512
Views:	1675
Downloads:	233
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Soil-moisture estimation from SAR images
Secondary abstract:	This dissertation presents soil-moisture estimation from Synthetic Aperture Radar (SAR) images, where backscattering coefficient is transformed into relative dielectric constant from which it is pos-sible to retrieve volumetricsoil-moisture values. The novelty of this dissertation is a development of a new soil-moisture parameter retrieval model. The proposed model works at X-band microwave radiation, and is composed from bare-soil estimation with additional correction part for small vegeta-tion. The soil-moisture estimation algorithm is compared with in-situ measurements collected simulta-neously to image acquisitions. For this dissertation high-resolution spot-light TerraSAR-X images were collected under different radar parameters, which include: various incidence angles, polarizations and range bandwidths. In this dissertation two different soil moisture estimation models are proposed, where the first one is based on the Shi semi-empirical model, thus it only works on bare-soil surfaces. The Shi semi-empirical model is modified in a sense, that new polynomials are introduced in the Shi model equation. Polynomial parameters are later determined using real synthetic aperture radar images represented as backscattering coefficients and in-situ measurement values. The second proposed model is more general model, because it is able to estimate soil-moisture from areas with small vegetation and is based on the water-cloud model. The soil-moisture is not only dependent on radar parameters, but also on soil surface roughness, that can be described with surface RMS height, correlation length, and correlationćs function. The other unknown parameters are vegetation fractional part and Fresnelćs reflectivity coefficient. The idea of the proposed model is that it is possible to estimate unknown parameters by using good properties of data restoration techniques, for which this dissertation relies on the Tikhonov regularization. This regularization estimates the unknown soil moisture parameters, where regularization parameters should cancel out or compensate the correla-tion length and correlationćs function. Experimental results showed that the proposed models provide good results for estimating soil-moisture on bare-soils and areas with small vegetation. However, the results show far superior estimation accuracy for the proposed two component model based on water-cloud model with Tikhonov regularization scheme.
Secondary keywords:	soil-moisture estimation;TerraSAR-X satellite;Tikhonov regularization;empirical model;
URN:	URN:SI:UM:
Type (COBISS):	Dissertation
Thesis comment:	Univ. Maribor, Fak. za elektrotehniko, računalništvo in informatiko
Pages:	XVI, 97 str., [12] str. pril.
Keywords (UDC):	science and knowledge;organization;computer science;information;documentation;librarianship;institutions;publications;znanost in znanje;organizacije;informacije;dokumentacija;bibliotekarstvo;institucije;publikacije;prolegomena;fundamentals of knowledge and culture;propaedeutics;prolegomena;splošne osnove znanosti in kulture;computer science and technology;computing;data processing;računalniška znanost in tehnologija;računalništvo;obdelava podatkov;
ID:	1026262