Geographical Weighted Regression Model For Improved Nearshore Water Depth Estimation From Multispectral Imagery
Video in TIB AVPortal:
Geographical Weighted Regression Model For Improved Nearshore Water Depth Estimation From Multispectral Imagery
Formal Metadata
Title 
Geographical Weighted Regression Model For Improved Nearshore Water Depth Estimation From Multispectral Imagery

Title of Series  
Author 

License 
CC Attribution  NonCommercial  ShareAlike 3.0 Germany:
You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal and noncommercial purpose as long as the work is attributed to the author in the manner specified by the author or licensor and the work or content is shared also in adapted form only under the conditions of this license. 
Identifiers 

Publisher 

Release Date 
2015

Language 
English

Producer 

Production Year 
2015

Production Place 
Seoul, South Korea

Content Metadata
Subject Area  
Abstract 
There is often a need for making a highresolution or a complete bathymetric map based on sparse point measurements of water depth. The common practice of previous studies has been to calibrate a single global depth regression model for an entire image. The performance of conventional global models is limited when the bottom type and water quality vary spatially within the scene .For a more accurate and robust waterdepth mapping , this study proposes a regression model for a geographical region or local area rather than using a global regression model. The global regression model and Geographical Weighted Regression (GWR) model are applied to Landsat 8 and RapidEye satellite images. The entire data analysis workflow was carried out using GRASS GIS Version 7.0.0. Comparison of results indicates that the GWR model improves the depth estimation significantly, irrespective of the spatial resolution of the data processed. GWR is also seen to be effective in addressing the problem introduced by heterogeneity of the bottom type and provide better bathymetric estimates in near coastal waters. The study was carried out at Pureto Rico, northeastern Caribbean sea. Two different satellite data were collected in order to test the algorithm with high and moderate resolution data. RapidEye data has 12bit radiometric resolution and 5 meter spatial resolution. Even though Landsat 8 data also has 12bit radiometric resolution, it provides 30 m spatial resolution. In order to calibrate and evaluate the estimated depth, high accuracy LiDAR depth data (4 m resolution) provided by NOAA is used. The study was demonstrating GWR model to estimate depth, evaluate and compare the results with a global conventional regression model. The comparative study between conventional global model and GWR model shows that GWR model significantly increases the accuracy of the depth estimates and addresses spatial heterogeneity issue of the bottom type and water quality. The GWR model provide better accuracy at both Landsat 8 (Rsquared=0.96 and RMSE=1.37m) and RapidEye (Rsquared=0.95 and RMSE=1.63m) than global model at Landsat 8 (Rsquared=0.71 and RMSE=3.71m) and RapidEye (Rsquared=0.71 and RMSE=4.04m).

00:00
Medical imaging
Algorithm
Estimator
Computer animation
Universe (mathematics)
Mathematical analysis
Event horizon
00:52
Regression analysis
Presentation of a group
Multiplication sign
Moment (mathematics)
Genetic programming
Interactive television
Set (mathematics)
Inequality (mathematics)
Mathematical model
Computer programming
Estimator
Computer animation
Lecture/Conference
Personal digital assistant
Operator (mathematics)
God
03:05
Group action
Greatest element
Functional (mathematics)
Algorithm
Electric generator
Connectivity (graph theory)
Reflection (mathematics)
Paralleler Algorithmus
Mathematical analysis
Water vapor
Maxima and minima
Mathematical model
Spektrum <Mathematik>
Element (mathematics)
Medical imaging
Computer animation
Internetworking
Singleprecision floatingpoint format
Universe (mathematics)
Order (biology)
Summierbarkeit
Musical ensemble
06:05
Area
Medical imaging
Dynamical system
Computer animation
Observational study
Cellular automaton
Software developer
Open set
Surface of revolution
07:16
Group action
Existence
Reflection (mathematics)
Angular resolution
Shared memory
Shape (magazine)
Event horizon
Mathematical model
Mathematical model
Goodness of fit
Estimator
Process (computing)
Computer animation
Term (mathematics)
Different (Kate Ryan album)
Energy level
Metropolitan area network
09:29
Area
Matching (graph theory)
Mapping
Cluster analysis
Genetic programming
Set (mathematics)
Mathematical model
Subset
Estimator
Message passing
Computer animation
Singleprecision floatingpoint format
Coefficient
Position operator
10:27
Estimator
Message passing
Computer animation
Mapping
Cluster analysis
Social class
10:56
Estimator
Graph (mathematics)
Computer animation
Term (mathematics)
Network topology
Singleprecision floatingpoint format
Set (mathematics)
Energy level
Musical ensemble
Electronic signature
Social class
11:44
Point (geometry)
Regression analysis
Pixel
Divisor
Execution unit
Student's ttest
Parameter (computer programming)
Mathematical model
Number
Medical imaging
Crosscorrelation
Reduction of order
Metropolitan area network
Linear regression
Structural load
Cellular automaton
Mathematical analysis
Instance (computer science)
Data management
Kernel (computing)
Computer animation
Personal digital assistant
Pearson productmoment correlation coefficient
Quotient
Coefficient
14:21
Standard deviation
Computer animation
Different (Kate Ryan album)
Personal digital assistant
Term (mathematics)
Operator (mathematics)
3 (number)
Incidence algebra
Food energy
Event horizon
Metropolitan area network
Physical system
15:42
Time zone
Category of being
Computer animation
Multiplication sign
Shared memory
Bit
Student's ttest
Line (geometry)
Mathematical model
Social class
God
16:52
Metre
Standard deviation
Computer animation
Incidence algebra
Limit (category theory)
Fundamental theorem of algebra
Resultant
Number
17:38
Area
Cross section (physics)
Computer animation
Plotter
Coefficient
Mathematical model
Computer programming
Social class
18:35
Axiom of choice
Sparse matrix
Computer animation
Observational study
Plotter
Mathematical model
Scattering
19:00
Algorithm
Estimator
Presentation of a group
Kernel (computing)
Computer animation
Observational study
Concentric
Internetworking
1 (number)
Mathematical model
Mathematical optimization
20:18
Word
Radius
Computer animation
Wellformed formula
Website
Form (programming)
Diameter
Number
21:37
Lecture/Conference
00:00
but how Lieutenant my name is
00:10
annotated candidate from us this at universities event so basic analysis is about the estimating notion that emitted from remote sensing it is basically optical remote sensing images so they're testing something algorithms and in growing up the algorithms and modifying modifying the existing algorithms and here we will talk about the the later modern which is using to estimate and no other that this vesicle using for all kind of spatially it violated estimation it is the aggression so this is this is the of
00:56
this is the moment of my presentation today in the interaction by the local refuted the border of the hollow optical remote sensing can be used for non showed up destination and and that the use of using them optical remote sensing for that destination and the physical principle behind that and we can use and at the end of the Lord the collection that there's this have been used to to cut the atom bomb cities already then the spending in the that lady is split should be removed for better estimation then that global regression method which is a conventional meant that it is a single symbol might win regression that the it is used by many others to estimate that and then that the president it is gradually so if I was a a those cells and compare as uh in about the ground that is collected so that means that the set the
02:03
book uh tags its to conduct to from the national operations so that is laid out as born later met imagery and the other 1 is nice owning so that meant that the occurs solution that does it but they're about the same the advantages because due to inaccessibility of such kind of big data and the time consuming and then it is of quiz questions answered program so there availability of some some data sets of God that it's on there and some of them are free and not all and that just go is getting good solution it does not have a solution it is only 900 of the solution I don't think it can be used for the case of cost and dynamic modelling so we must have so many others where claimed to have it the media due that that you see so all
03:07
that fundamentally the have been used that multispectral images so there are lots of advantages and the sum of some of somebody certain it is also that the universe is that the wait element of bad it's like somebody said let images lots of bands can be used to estimate effectively many 11 in quality and things like that it is a that in the course and then you come back to that the collections and last get of college and high spatial solutions and as a member of that accessibility and occlusive of sadly demands that you can have and convert that imitate is to limit the evidence here it is the tendency of group and the other quality made the water quality the stuff how how can the late contentedly to the bottom of the wonder if interest of cross Gosselin them and a lot of is very complex so it's not easy to bed up to the bottom so the use of the Internet and disease that really it is related to the accuracy and compared with the other might the goes on in the and not what you that this is really have to collect a lot of that led can to come there are advantages and so this is the
04:37
physical principle of their the algorithm this is actually a generation of light in water is a function of the interview should take a shot lakes In the electromagnetic spectrum it's candidate of maximum water there it is written here hunt made of which is not order the I cannot be uploading a accustomed complex model which can be written as the that the or awesome image something and if you know when you're increases your activation speed decreases so a single motivated parallel and long valence of the components of the Roanoke electromagnetic spectrum it's an authenticated off water water meters I do next from the Shaolin order it's an so I don't consider Europe that the firm and sometimes it is kind of a centered that the when we met reflectance and release the neural this is 1 all of these in the name think it's the water and you for have formed within the reach of water and and by avoiding analysis that was the excess salt among my mind this is only 1 of the things I have to remove all of the of the other components so this
06:06
is the study area give the entities ladies and understand that in the back to the user to select but do you think is the study of how it is that a few and has a a solution that that that that it could be used for calibration and then there an of those cells and I have always been a which is a common belief in it gives you a sense that led to the development of that it's going to rapidly and Indian sometimes and don't the associated to a of and it decided that efforts the most
06:52
interesting and and that I have got the tools tools that let images when this land settled which is free of the Open Data and everything it says how special that was innocent and had a lunatic the solution of the term in dynamic and then the and the advantages of social revolution yes so
07:19
that if you have the but if you think about the connection that that I have been used to furthermore is the inexistent Edward Said have no modifications in terms of in intensified that induces lands and that the there are there to new events available can be demands at 7 which is going to cost event the venues for estimation because it it it is very shot Maryland get invented to it and Sharon Gillett advantages this is an higher the the the bad and this man is the use of our the action and other events o was involved estimate these this has been so for estimation and sharing and then using for that action because Schaeffer in shape invalid can have been added alot of model is that investment than it is to the book Madison and all that kind of good about shared interested and on the and the reflectance and so you can use evidence to the things that and that an abundance from the it's so this is that have that that
08:43
at the models are aggressive model is a conventional out of it is being used to raise some money those and this just giving workers engine the the but you really good resolution process but the bad then we examine the social again found that it is not intended to it that does that had to do it in the that it is due to the fact that and diets and different level quality especially special problems of having but and that's and different rather quality so really I was thinking of having an and many of which can be effectively and that's the problem due to the special heterogeneity the sale this is the 1st
09:34
German edition on the left side and this match is the bank the subset investment in that debate rather than position on it and and from that you that you know that there are also showing that the that suggests that not uniformly distributed on the map and it is it has a clusters of messages that that means that some of the local in US have influence on estimating and before going to
10:07
tell you 1 thing yeah navigational it is a symbol notable have Gaussian model which is having a single set of coefficients which is the reason is that of course this and then let live and that all that area so it is then is bad and so
10:28
that the me that you have to have sensitive coefficient estimate about the out of the data the so out out
10:41
here is this as the special the messages clusters to produce classification map so that led to that class that can be used for better but the the 1st emissions so I'm going as
10:59
tree so you have to have 1st IMAX
11:02
electoral estimates regardless flow in this usable as classification vacation to use testing the but the glasses and the signature we have selected the level where there where the terms in those regions and the graph here that that of problem that and the xaxis transcended his band and the leader dead and again so that needs into classes each having different paper uh scattering so that means that could be effectively at this by having a different gradations instead of having a single set of quotation so after that after a
11:50
student I have saying there was just happen again but the plasma is managed it is better than the performance its use and distributed and some of the that problem but it is that and then the parameters and correlation coefficient and essentially and determine the efficiency of the agent at the Efficient but the plasma is no it isn't in C n open and vocalization it is and
12:22
keep in mind that but I would have in the introduced it is not it doesn't exist in the is this a 1st and then based on regression analysis and it is called you the blue model geographical data and in case of abnormalities in they don't know the linear regression that is addressed this man aggression so the here the image kernel like this and it's gonna in the kernel you that is invertible and you estimate coefficient for this and other point by by validating your Anderson points if you the from this point will come as more knowledge for this point and if you go and from the center point of the load units so that if you have a special quotient but you had a good good correlation from the especially especially that the delegation so that it is the main factor here if you have not been so that is it's the bank of Canada yes I mean it's not going small so at the expense of the accuracy of the cells in you lab 1 if you if you are not going to do that and to can instances it is sparse then this this number of then new can be the is of the dialog manager of the Kennedy big an agent and what will that be it cannot be good solutions next not very significant i reduction analysis and this is good used to evade the each it's point and I you so that they do the crowd as estimated coefficients for each pixel in that up so that you have that for image pixels the challenge is mandatory the this so in 2008
14:26
and have this man difference in years to come back those cells and in that the now you have focused the energy of the incident to adapted to estimate the category that that in that case it can get have lot use insisted on to estimate something you will get lower solution and that the government had not good good dense that data and it did it do that and fascinating and I'm just a 3rd of the and that it affects the operation so you get this kind of standards and best and in sense so
15:03
the 1st thing you know the random and that the advantages of hands and didn't have has then and that was the case of that the system that inspections the and the imagination is kind of different events in the business to estimate the there and television shows and it is a now we know that says that the Americans in in the significant terms of analyzing and has that move
15:43
and this is then this category of further it's matter it is as shown that did you Due to the model is far better than the delegation out the profit sharing and the has the end there is no doubt about that and you know what that the semantic classes and the model is little bit at a time and time and time differences in see if you and
16:18
it has the 2nd sentence you have certain lines of infected persons just the at U. S examine them the presence of God I and in the middle of the inspectors tend to have the matter in the so that but the navigation model of students but it in the and the and and considered him a disadvantage because at the 2nd so student has
16:56
no fundamentalisms connected to the traveling incentive to so this
17:05
that there are some of its intentions to uh at 600 meter in the and the 2 thousand and the same kind of sentiment of results and the main thing is on the 1st of 2 persons and 1st of a standard that and this further damaging broadens its little because the course syllabus and 14 data for the limit of land so 1st of Canada is that even less number of incident that in the the um but at that means that this initially so this has
17:41
developed scatterplot so showing same kind in the way it scatter plot and crosssection professors and you
17:51
extrapolated destination because if if you have a smaller it out in support of the new underestimate followed up on that seemed to have a coastal area that in and the classes for all aerial that an inch classes the standard coefficients from these programs and that coefficients happen when the other areas so that is something that we can know about model which is something I didn't have work the model and you that that have different different coefficients from so that that that I can see them depending on tha
18:37
so this is the scatter plot of the study and then the candidate grasp of
18:46
it is German uh shoppers precedent sparse dumped open 5 kilometer and this preference as a showing this as an the Gw model choice
19:02
and the current concentration the is that doesn't the dynamical in this of rapid eye and land said a data can be effectively used for would estimation of that from the shock and gobbled aggression modeling is there is not they able to and does that it was unity after that done by June when the jitter be again I does the problem affected and here after the algorithm that have been done that indeed did the created as a module another in the future studies thank you
19:44
FIL and the the this the and no piston an internet Iliescu piston I'm on of his come member book of get go OK so I'm a highly into your presentation they're on all why these so optimum size or optimum Paul van Meegeren or you cannot kernel can I guess it is depends upon them the uh it is depends
20:22
upon your that Institute of this this is you have a new user for calibration if you have word then so that justice you can have not now the all of this that it 15 50 not in the middle of the fixes and so 1 the the city the fighting the site that means that the radius of the government and if you have large number of letters in this 10 thousand boys have used I have lots of ways to hear lot than the review the reduced I think is all you're a you're phrases sheet that can be converted to any of formula are awesome we regret a simple word as emperor our formula then might be created because of the IEEE feed youth answered the the diameter of the sugar meter OK so there should be some all the that many the relationship the this is the I'm may come in is there any form it no more questions at the they giving you think you the