© July 2011 Linear and Nonlinear Imaging Spectrometer Denoising Algorithms Assessed Through Chemistry Estimation David G. Goodenough 1,2, Geoffrey S. Quinn.

Slides:



Advertisements
Similar presentations
Digital Image Processing
Advertisements

REQUIRING A SPATIAL REFERENCE THE: NEED FOR RECTIFICATION.
Predicting and mapping biomass using remote sensing and GIS techniques; a case of sugarcane in Mumias Kenya Odhiambo J.O, Wayumba G, Inima A, Omuto C.T,
Natural Resources Canada Ressources naturelles Canada Canadian Forest Service Service canadien des forêts Conseil national de recherches Canada National.
Major Operations of Digital Image Processing (DIP) Image Quality Assessment Radiometric Correction Geometric Correction Image Classification Introduction.
Airborne Laser Scanning: Remote Sensing with LiDAR.
Radiometric and Geometric Errors
ASTER image – one of the fastest changing places in the U.S. Where??
Digital Imaging and Remote Sensing Laboratory Correction of Geometric Distortions in Line Scanner Imagery Peter Kopacz Dr. John Schott Bryce Nordgren Scott.
Principal Component Analysis
Hyperspectral Imagery
Questions How do different methods of calculating LAI compare? Does varying Leaf mass per area (LMA) with height affect LAI estimates? LAI can be calculated.
Airborne LIDAR The Technology Slides adapted from a talk given by Mike Renslow - Spencer B. Gross, Inc. Frank L.Scarpace Professor Environmental Remote.
Digital Imaging and Remote Sensing Laboratory Real-World Stepwise Spectral Unmixing Daniel Newland Dr. John Schott Digital Imaging and Remote Sensing Laboratory.
Comparison of LIDAR Derived Data to Traditional Photogrammetric Mapping David Veneziano Dr. Reginald Souleyrette Dr. Shauna Hallmark GIS-T 2002 August.
SPECTRAL AND HYPERSPECTRAL INSPECTION OF BEEF AGEING STATE FERENC FIRTHA, ANITA JASPER, LÁSZLÓ FRIEDRICH Corvinus University of Budapest, Faculty of Food.
UNDERSTANDING LIDAR LIGHT DETECTION AND RANGING LIDAR is a remote sensing technique that can measure the distance to objects on and above the ground surface.
Chapter 12 Spatial Sharpening of Spectral Image Data.
Mapping Forest Vegetation Structure in the National Capital Region using LiDAR Data and Analysis Geoff Sanders, Data Manager Mark Lehman, GIS Specialist.
An overview of Lidar remote sensing of forests C. Véga French Institute of Pondicherry.
1 Image Pre-Processing. 2 Digital Image Processing The process of extracting information from digital images obtained from satellites Information regarding.
Remote Sensing Hyperspectral Remote Sensing. 1. Hyperspectral Remote Sensing ► Collects image data in many narrow contiguous spectral bands through the.
Noise-Robust Spatial Preprocessing Prior to Endmember Extraction from Hyperspectral Data Gabriel Martín, Maciel Zortea and Antonio Plaza Hyperspectral.
Chenghai Yang 1 John Goolsby 1 James Everitt 1 Qian Du 2 1 USDA-ARS, Weslaco, Texas 2 Mississippi State University Applying Spectral Unmixing and Support.
ARSF Data Processing Consequences of the Airborne Processing Library Mark Warren Plymouth Marine Laboratory, Plymouth, UK RSPSoc 2012 – Greenwich, London.
A New Subspace Approach for Supervised Hyperspectral Image Classification Jun Li 1,2, José M. Bioucas-Dias 2 and Antonio Plaza 1 1 Hyperspectral Computing.
Blue: Histogram of normalised deviation from “true” value; Red: Gaussian fit to histogram Presented at ESA Hyperspectral Workshop 2010, March 16-19, Frascati,
Orthorectification using
Quality Assessment for LIDAR Point Cloud Registration using In-Situ Conjugate Features Jen-Yu Han 1, Hui-Ping Tserng 1, Chih-Ting Lin 2 1 Department of.
Image Preprocessing: Geometric Correction Image Preprocessing: Geometric Correction Jensen, 2003 John R. Jensen Department of Geography University of South.
Part 1: Basic Principle of Measurements
Fuzzy Entropy based feature selection for classification of hyperspectral data Mahesh Pal Department of Civil Engineering National Institute of Technology.
1 Exploiting Multisensor Spectral Data to Improve Crop Residue Cover Estimates for Management of Agricultural Water Quality Magda S. Galloza 1, Melba M.
Digital Imaging and Remote Sensing Laboratory NAPC 1 Noise Adjusted Principal Component Transform (NAPC) Data are first preprocessed to remove system bias.
Generating fine resolution leaf area index maps for boreal forests of Finland Janne Heiskanen, Miina Rautiainen, Lauri Korhonen,
Compression and Analysis of Very Large Imagery Data Sets Using Spatial Statistics James A. Shine George Mason University and US Army Topographic Engineering.
A Study on the Effect of Spectral Signature Enhancement in Hyperspectral Image Unmixing UNDERGRADUATE RESEARCH Student: Ms. Enid Marie Alvira-Concepción.
VEGETATION Narrow- vs. Broad-Band Instruments Wavelength (nm) Reflectance TM Bands.
Chapter 8 Remote Sensing & GIS Integration. Basics EM spectrum: fig p. 268 reflected emitted detection film sensor atmospheric attenuation.
RASTERTIN. What is LiDAR? LiDAR = Light Detection And Ranging Active form of remote sensing measuring distance to target surfaces using narrow beams of.
Digital Image Processing Definition: Computer-based manipulation and interpretation of digital images.
Change Detection The process of identifying differences in the state of an object or phenomenon by observing it at different times. Change detection applications:
ISOMAP TRACKING WITH PARTICLE FILTER Presented by Nikhil Rane.
14 ARM Science Team Meeting, Albuquerque, NM, March 21-26, 2004 Canada Centre for Remote Sensing - Centre canadien de télédétection Geomatics Canada Natural.
Colorado Center for Astrodynamics Research The University of Colorado 1 STATISTICAL ORBIT DETERMINATION The Minimum Variance Estimate ASEN 5070 LECTURE.
A NOVEL METHOD FOR COLOR FACE RECOGNITION USING KNN CLASSIFIER
Hyperspectral remote sensing
Pixel Clustering and Hyperspectral Image Segmentation for Ocean Colour Remote Sensing Xuexing Zeng 1, Jinchang Ren 1, David Mckee 2 Samantha Lavender 3.
Digital Imaging and Remote Sensing Laboratory Maximum Noise Fraction Transform 1 Generate noise covariance matrix Use this to decorrelate (using principle.
2011 IEEE International Geoscience and Remote Sensing Symposium (IGARSS) Aihua Li Yanchen Bo
SGM as an Affordable Alternative to LiDAR
U NIVERSITY OF J OENSUU F ACULTY OF F ORESTRY Introduction to Lidar and Airborne Laser Scanning Petteri Packalén Kärkihankkeen ”Multi-scale Geospatial.
Detection of Wind Speed and Sea Ice Motion in the Marginal Ice Zone from RADARSAT-2 Images Alexander S. Komarov 1, Vladimir Zabeline 2, and David G. Barber.
Shadow Detection in Remotely Sensed Images Based on Self-Adaptive Feature Selection Jiahang Liu, Tao Fang, and Deren Li IEEE TRANSACTIONS ON GEOSCIENCE.
European Geosciences Union General Assembly 2016 Comparison Of High Resolution Terrestrial Laser Scanning And Terrestrial Photogrammetry For Modeling Applications.
Integrating LiDAR Intensity and Elevation Data for Terrain Characterization in a Forested Area Cheng Wang and Nancy F. Glenn IEEE GEOSCIENCE AND REMOTE.
Counting the trees in the forest
PADMA ALEKHYA V V L, SURAJ REDDY R, RAJASHEKAR G & JHA C S
Background on Classification
Factsheet # 19 Understanding multiscale dynamics of landscape change through the application of remote sensing & GIS Hyperspectral Remote Sensing of Urban.
Digital Data Format and Storage
Orthogonal Subspace Projection - Matched Filter
Why LiDAR makes hyperspectral imagery more valuable for forest species mapping OLI 2018 Andrew Brenner, Scott Nowicki & Zack Raymer.
Hyperspectral Image preprocessing
By: Paul A. Pellissier, Scott V. Ollinger, Lucie C. Lepine
Satellite data Marco Puts
GAUSSIAN PROCESS REGRESSION WITHIN AN ACTIVE LEARNING SCHEME
Somi Jacob and Christian Bach
Spectral Transformation
2011 International Geoscience & Remote Sensing Symposium
Presentation transcript:

© July 2011 Linear and Nonlinear Imaging Spectrometer Denoising Algorithms Assessed Through Chemistry Estimation David G. Goodenough 1,2, Geoffrey S. Quinn 3, Piper L. Gordon 2, K. Olaf Niemann 3 and Hao Chen 1 1 Pacific Forestry Centre, Natural Resources Canada, Victoria, BC 2 Department of Computer Science, University of Victoria, Victoria, BC 3 Department of Geography, University of Victoria, Victoria, BC

© July 2011 Linear and Nonlinear Denoising Algorithms Assessed Through Chemistry Estimation  Objective: To compare linear and non-linear methods of denoising hyperspectral data; do we always need non-linear methods?  Data collection: Study area, sample collection, data/sensor characteristics  Pre-processing: Orthorectification and radiometric calibration  Processing: Contextual filter, spectral transformations, PLS regression, Chlorophyll-a and Nitrogen estimation  Analysis:  30 x 30 m Plot-level  2 x 2 m Tree-level  Conclusions

© July 2011 Data collection: The Greater Victoria Watershed District (GVWD) 14 plots, 140 trees

© July 2011  Acquisition date  September 11, 2006  Spectral data  Range: nm  492 spectral bands  Mean sampling interval: 2.37nm (VNIR <990nm) 6.30nm (SWIR>1001)  Mean FWHM: 2.37nm (VNIR) 6.28 (SWIR)  Spatial data  300 spatial pixels  FOV: 22°  IFOV: 0.076°  Imaging rate: 40f/s  Flight speed: 70m/s  Along track sampling: 1.75m  Flight altitude: 1500m  2m resolution Data collection: AISA Hyperspectral Data Acquisition

© July 2011  Sensor characteristics  Discrete return LIDAR system  1064 nm  FOV: 20°  Footprint: ~25 cm (variable)  Pulse rate: 100+ Khz  Scan rate: 15 to 30 Hz  Flight speed: 70 m/s  Flight altitude: 1500m  Posting density: ~1.2/m 2  Data  Applanix 410 IMU/DGPS system  First and last return x, y, z positions  Range accuracy: 5 to 10 cm  Rasterized to 2m resolution corresponding to AISA data  Canopy height, digital surface and bare earth models are derived  Acquisition date  Concurrent with AISA acquisition Data collection: Lidar Data Acquisition

© July 2011  Geometric distortions (non-uniform distance and direction) caused by platform altitude, attitude (roll, pitch and yaw) and surface relief  Traditional DEM orthorectification at fine resolutions introduce significant errors in tree canopy positions  Accurate positioning is vital for high resolution datasets and fine scale patterns and processes  The lidar RBO (range based orthorectification), reduces misregistration issues caused by layover of the reflected surface.  Atmospheric corrections performed by ATCOR-4 (airborne) software applying sensor and atmospheric parameters to sample MODTRAN LUT and provide correction factors  Empirical line calibration performed to reduce residual errors AISA (B,G,R: 460,550,640nm) draped over LIDAR DSM Data pre-processing: Radiometric and Geometric Correction

© July 2011 Nonlinearity of Hyperspectral  Hyperspectral data is non-linear  Minimum Noise Fraction (MNF)  Popular linear noise removal technique  Non-linear Local Geometric Projection Algorithm (NL-LGP)  Will it outperform MNF denoising for foliar chemistry prediction? T. Han and D. G. Goodenough, "Investigation of Nonlinearity in Hyperspectral Imagery Using Surrogate Data Methods," Geoscience and Remote Sensing, IEEE Transactions on, vol. 46, pp , 2008.

© July 2011 Denoising: Linear and Nonlinear AISA image 180 m x 170 m area True colour RGB: 1736, 1303, 1089nm Difference Images Inverse MNF denoisedNL-LGP denoised NL-LGP - ReflectanceReflectance - MNF

© July 2011 NL-LGP Algorithm  Construct state vectors in phase space  Specify the neighbourhood of these state vectors  Find projection directions  Project the state vectors on these directions, reducing noise D. G. Goodenough, H. Tian, B. Moa, K. Lang, C. Hao, A. Dhaliwal, and A. Richardson, "A framework for efficiently parallelizing nonlinear noise reduction algorithm," in Geoscience and Remote Sensing Symposium (IGARSS), 2010 IEEE International, pp

© July 2011 Minimum Noise Fraction  Estimates noise in the data and in a Principal Components Analysis (PCA) of the noise covariance matrix  Noise whitening models the noise in the data as having unit variance and being spectrally uncorrelated  A second PCA is taken  Resulting MNF eigenvectors are ordered from highest to lowest signal to noise ratio (noise variance divided by total variance)

© July 2011 Plot-Level Chemistry Comparison Process AISA 30m data AISA 2m data MNF denoised data NL-LGP denoised data Averaging Inverse MNF denoising NL-LGP denoising Reflectance chemistry predictions MNF denoised chemistry predictions NL-LGP denoised chemistry predictions Chemistry ground data Partial Least Squares (PLS) Regression PLS Regression PLS Regression

© July 2011 Spectral Transformation for Comparing Chemistry Predictions  Mean R 2 values for the transformation types are output by the PLS program  Large standard deviations, overlapping between original reflectance, MNF and NL-LGP denoised  2 nd derivative (2 points left) has one of the highest mean R-squared values  The most accurate predictions from PLS regression are output for each transformation type  2 nd derivative (2 points left) gave best prediction for all 3 spectra types and both Nitrogen and Chlorophyll-a chemistry

© July 2011 Plot-Level Average R-squared Values for Nitrogen

© July 2011 Plot-Level Non-current Nitrogen (% dry weight)

© July 2011 PLS Plot-Level Chlorophyll-a (μg/mg)

© July 2011 Moving from Plot-Level to Tree-Level  Original reflectance predicts chemistry with greater accuracy than denoised reflectance  Averaging from 2 x 2 m pixels to 30 x 30 m pixels  Preprocessing of the data (orthorectification and radiometric calibration)  To find if there is non-linear noise at the 2 m level (tree-level) the process is repeated with original, non- averaged AISA 2 m data

© July 2011 Tree-Level Chemistry Comparison Process AISA 2m data MNF denoised data NL-LGP denoised data Inverse MNF denoising NL-LGP denoising Reflectance chemistry predictions MNF denoised chemistry predictions NL-LGP denoised chemistry predictions Chemistry ground data Partial Least Squares (PLS) Regression PLS Regression PLS Regression

© July 2011 Tree-Level Chemical Analysis  Spectra were extracted from the positions of each tree in the plot data (2m by 2m pixels)  Chemistry predictions were generated for the ten trees in each of the 14 plots, against the averaged chemistry measurement for their plot  2 nd derivative of reflectance (2 points left) gave the best R 2 values and was used for the chemistry predictions

© July 2011 Tree-Level Chemistry Comparison 14 Plots 140 Trees Predicted Chemistry for each of… MNF denoised NL-LGP denoised AISA 2m reflectance Averaged Measured Chemistry vs

© July 2011 PLS Tree-Level Non-current Nitrogen (% dry weight)

© July 2011 PLS Tree-Level Chlorophyll-a (μg/mg)

© July 2011 Conclusions: Linear and Non-Linear Denoising Algorithms  For plot-level applications, denoising is not necessary  The averaging process is effective for removing noise  For tree-level applications, use of a non-linear denoising method is better for mapping chemistry  Nitrogen  Non-Linear ±  MNF ±  Original Reflectance ±  Chlorophyll  Non-Linear ±  MNF ±  Original Reflectance ± 0.054

© July 2011 Conclusions: Linear and Non-Linear Denoising Algorithms  MNF does not improve chemistry predictions, further supporting the non-linearity of hyperspectral data  The application of PLS regression to forest chemistry mapping remains our most reliable method for chemistry estimation  R 2 of ~0.9 for plot-level  R 2 of ~0.8 for tree-level

© July 2011 We thank: The University of Victoria for its support. Natural Resources Canada (NRCan), the Canadian Space Agency (CSA), and Natural Sciences and Engineering Research Council of Canada (NSERC) (DGG) for their support. The Victoria Capital Regional District Watershed Protection Division for its logistical support. The audience for their attention. Acknowledgements: Hyperspectral applications for forestry