A Comparison of Land Use and Land Cover Change Detection Methods Daniel L. Civco, James D. Hurd, Emily H. Wilson, Mingjun Song, Zhenkui Zhang C enter for L and use E ducation A nd R esearch Department of Natural Resources Management & Engineering The University of Connecticut U-4087, Room 308, 1376 Storrs Road Storrs, CT
OutlineOutline BackgroundBackground ObjectivesObjectives Study Area and DataStudy Area and Data MethodsMethods –Post Classification Analysis –Cross-correlation Analysis –Neural Networks –Segmentation & Object-oriented Classification ResultsResults ConclusionsConclusions RecommendationsRecommendations BackgroundBackground ObjectivesObjectives Study Area and DataStudy Area and Data MethodsMethods –Post Classification Analysis –Cross-correlation Analysis –Neural Networks –Segmentation & Object-oriented Classification ResultsResults ConclusionsConclusions RecommendationsRecommendations
Northeast Applications of Useable Technology In Land planning for Urban Sprawl A NASA Regional Earth Science Applications Center (RESAC) A NASA Regional Earth Science Applications Center (RESAC)
To educate the general public on the value and utility of geospatial technologies, particularly RS information. Our RESAC Mission To make the power of remote sensing technology available, accessible and useable to local land use decision makers as they plan their communities.
NAUTILUS Research Better land cover mapping and change detection Urban growth models and metrics Forest fragmentation models and metrics Improved impervious cover estimates
Need for effective methods for deriving information on –Land use change –Forest fragmentation –Urban growth –Loss of agricultural lands –Increase in impervious surface area Need for effective methods for deriving information on –Land use change –Forest fragmentation –Urban growth –Loss of agricultural lands –Increase in impervious surface area BackgroundBackground
BackgroundBackground Farmland conversion in the Stony Brook Millstone Watershed: Genesis 3D rendering of land use derived from Landsat data
ObjectiveObjective Compare the results of different land use and land cover change detection approaches traditional post-classification cross-tabulation cross-correlation analysis neural networks knowledge-based expert systems image segmentation & object- oriented classification Compare the results of different land use and land cover change detection approaches traditional post-classification cross-tabulation cross-correlation analysis neural networks knowledge-based expert systems image segmentation & object- oriented classification
Research & Education Watersheds A range of land covers and issues Presumpscot SuAsCo Salmon Stonybrook
Stony Brook Millstone Watershed, NJ USGS MRLC 265 Sq. Miles USGS MRLC 265 Sq. Miles Has a strong Watershed Association in existence Has a strong Watershed Association in existence Between New York City and Philadelphia Between New York City and Philadelphia Increased development pressures Loss of agriculture land to urban sprawl Loss of agriculture land to urban sprawl
Study Area and Data SITE 1 SITE 2 Stony Brook Millstone Watershed Stony Brook Millstone Watershed
March 27, 1989 May 4, 2000 Study Area and Data Site 1
September 3, 1989 September 23, 1999 Study Area and Data Site 1
March 27, 1989 May 4, 2000 Study Area and Data Site 2
September 3, 1989 September 23, 1999 Study Area and Data Site 2
Study Area and Data September 23, 1999 September 3, 1989
MethodsMethods –Post Classification Analysis –Cross-correlation Analysis –Neural Networks –Segmentation & Object-oriented Classification –Post Classification Analysis –Cross-correlation Analysis –Neural Networks –Segmentation & Object-oriented Classification
Post Classification Analysis Step 1 –Unsupervised classification Identify known clusters Extract unknown clusters Step 1 –Unsupervised classification Identify known clusters Extract unknown clusters 1989 Classification Iteration 1 Water ConiferousDeciduousAgriculture Turf & Grass Residential Dense Urban WetlandBarrenUnknown
Post Classification Analysis Step 2 –Unsupervised classification Identify known clusters Extract unknown clusters Step 2 –Unsupervised classification Identify known clusters Extract unknown clusters 1989 Classification Iteration 2 Water ConiferousDeciduousAgriculture Turf & Grass Residential Dense Urban WetlandBarrenUnknown
Post Classification Analysis Step 3 –Unsupervised classification Identify known clusters Extract unknown clusters Step 3 –Unsupervised classification Identify known clusters Extract unknown clusters 1989 Classification Iteration 3 Water ConiferousDeciduousAgriculture Turf & Grass Residential Dense Urban WetlandBarrenUnknown
Post Classification Analysis Step 4 –Unsupervised classification Identify clusters Step 4 –Unsupervised classification Identify clusters 1989 Classification Iteration 4 Water ConiferousDeciduousAgriculture Turf & Grass Residential Dense Urban WetlandBarrenUnknown
Post Classification Analysis Step 5 –Combine iterations into single land cover image Step 5 –Combine iterations into single land cover image Water ConiferousDeciduousAgriculture Turf & Grass Residential Dense Urban WetlandBarren
Post Classification Analysis Step 6 –Smooth image using majority filters Step 6 –Smooth image using majority filters Water ConiferousDeciduousAgriculture Turf & Grass Residential Dense Urban WetlandBarren
Post Classification Analysis 1989 Classification 2000 Classification Site 1 Perform similar procedure on 2000 date
Post Classification Analysis 1989 Classification 2000 Classification Site 2
Post Classification Analysis Cross Tabulate 1989 Classification 2000 Classification Change Site 2
MethodsMethods –Post Classification Analysis –Cross-correlation Analysis –Neural Networks –Segmentation & Object-oriented Classification
Cross-correlation Analysis CCA calculates the sum of the distance of each pixel in each band from the norm Z is the distance measure Observed is the pixel value for each band Expected is the mean value of all extracted pixels for each band Std. Dev. Is the standard deviation of all extracted pixels for each band
Cross-correlation Analysis Step 1 –Use 1989 classification as base land cover –Extract vegetated class areas to be analyzed from 1999/2000 ETM image (turf & grass, agriculture & barren, deciduous, and coniferous) Step 1 –Use 1989 classification as base land cover –Extract vegetated class areas to be analyzed from 1999/2000 ETM image (turf & grass, agriculture & barren, deciduous, and coniferous) September 23, Deciduous Category
Cross-correlation Analysis Step 2 –Perform CCA on 1999/2000 imagery –Identify thresholds separating unchanged pixels and changed pixels Step 2 –Perform CCA on 1999/2000 imagery –Identify thresholds separating unchanged pixels and changed pixels Probable unchanged Probable changed 1989 Deciduous Category Z-values range from 1 to 5,794 Z-values range from 1 to 5,794
Cross-correlation Analysis Step 3 –Create a mask from changed pixels for all categories analyzed (turf & grass, agriculture & barren, deciduous, and coniferous) Step 3 –Create a mask from changed pixels for all categories analyzed (turf & grass, agriculture & barren, deciduous, and coniferous) DeciduousConiferous Turf & GrassAgriculture & Barren
Cross-correlation Analysis Step 4 –Extract changed pixels from 1999/2000 image data –Perform unspervised classification to identify new categories Step 4 –Extract changed pixels from 1999/2000 image data –Perform unspervised classification to identify new categories Water ConiferousDeciduousAgriculture Turf & Grass Residential Dense Urban WetlandBarren
Cross-correlation Analysis Step 4 –Merge new classes with historic classification to produce updated land cover Step 4 –Merge new classes with historic classification to produce updated land cover Water ConiferousDeciduousAgriculture Turf & Grass Residential Dense Urban WetlandBarren
Cross-correlation Analysis 1989 Classification 2000 Classification Site 1
Cross-correlation Analysis 1989 Classification 2000 Classification Site 2
MethodsMethods –Post Classification Analysis –Cross-correlation Analysis –Neural Networks –Segmentation & Object-oriented Classification
Neural Networks N autilus I mage P rocessing S ystem
Neural Networks Step 1 –Select training features based on points Step 1 –Select training features based on points
Neural Networks Step 2 –Extract digital numbers for: Each pixel –By class Each Band Step 2 –Extract digital numbers for: Each pixel –By class Each Band
Neural Networks Records259 Features7 Classes9 Band 1Band 2Band 3Band 4Band 5Band 7Band 6Class … … …. Example of Site 2 Training Data for September 23, 1999 Class actually represented by one-of-n encoding. (i.e.,
Neural Networks Step 3: Create data set of all possible from T 1 to T 2 changes (constrained by permitted changes)
Neural Networks From To Urban ResidntlTurf&GrassAgricDecidConifWaterWetlandBarren Urban Residential Turf&Grass Agriculture Deciduous Coniferous Water Wetland Barren Permitted Changes
Neural Networks Step 4 –Create neural network classifier NeuralSIM ® –Backpropagation –Export C-code –Compile into NIPS Step 4 –Create neural network classifier NeuralSIM ® –Backpropagation –Export C-code –Compile into NIPS
Neural Networks Step 5 –Perform full neural network- based change detection within NIPS Step 5 –Perform full neural network- based change detection within NIPS
MethodsMethods –Post Classification Analysis –Cross-correlation Analysis –Neural Networks –Segmentation & Object-oriented Classification
Segmentation and Object- oriented Classification Step 1: Preprocessing –Prepare Image Data (2 dates and 2 seasons = 4 images) –Use indices to extract obvious classes –Create a data layer using the knowledge engineer –Add to image data for input into eCognition Step 1: Preprocessing –Prepare Image Data (2 dates and 2 seasons = 4 images) –Use indices to extract obvious classes –Create a data layer using the knowledge engineer –Add to image data for input into eCognition
Segmentation and Object- oriented Classification Step 1: Preprocessing –Prepare Image Data (2 dates and 2 seasons = 4 images) –Use indices to extract obvious classes –Create a data layer using the knowledge engineer –Add to image data for input into eCognition Step 1: Preprocessing –Prepare Image Data (2 dates and 2 seasons = 4 images) –Use indices to extract obvious classes –Create a data layer using the knowledge engineer –Add to image data for input into eCognition Date 1 Spring NDVI Date 1 Summer NDVI Date 2 Summer NDMI Date 2 Spring NDVI Date 1 Spring NDMI Date 1 Summer NDMI Date 2 Summer NDVI Date 2 Spring NDMI
Segmentation and Object- oriented Classification Step 1: Preprocessing –Prepare Image Data (2 dates and 2 seasons = 4 images) –Use indices to extract obvious classes –Create a data layer using the knowledge engineer –Add to image data for input into eCognition Step 1: Preprocessing –Prepare Image Data (2 dates and 2 seasons = 4 images) –Use indices to extract obvious classes –Create a data layer using the knowledge engineer –Add to image data for input into eCognition
Segmentation and Object- oriented Classification Step 1: Preprocessing –Prepare Image Data (2 dates and 2 seasons = 4 images) –Use indices to extract obvious classes –Create a data layer using the knowledge engineer –Add to image data for input into eCognition Step 1: Preprocessing –Prepare Image Data (2 dates and 2 seasons = 4 images) –Use indices to extract obvious classes –Create a data layer using the knowledge engineer –Add to image data for input into eCognition Site 1 Site 2
Segmentation and Object- oriented Classification Step 2 –Create eCognition Projects for date 1 and date 2 –Segment images using both seasons of imagery (excluding layer 7) Step 2 –Create eCognition Projects for date 1 and date 2 –Segment images using both seasons of imagery (excluding layer 7) Input Data: 7 layers 1. Red (spring) 2. NIR (spring) 3. MIR (spring) 4. Red (summer) 5. NIR (summer) 6. MIR (summer) 7. Classified layer
Segmentation and Object- oriented Classification Step 2 –Create eCognition Projects for date 1 and date 2 –Segment images using both seasons of imagery (excluding layer 7) Step 2 –Create eCognition Projects for date 1 and date 2 –Segment images using both seasons of imagery (excluding layer 7) Segment four levels LevelScaleColorShape
PixelLevel 1Level 2 Level 3 Level 4 Segmentation and Object- oriented Classification Step 2 –Create eCognition Projects for date 1 and date 2 –Segment images using both seasons of imagery (excluding layer 7) Step 2 –Create eCognition Projects for date 1 and date 2 –Segment images using both seasons of imagery (excluding layer 7)
Segmentation and Object- oriented Classification Step 3 –Begin creating class hierarchy –Training segment selection and standard nearest neighbor –Nearest Neighbor Classification of each level Step 3 –Begin creating class hierarchy –Training segment selection and standard nearest neighbor –Nearest Neighbor Classification of each level
Segmentation and Object- oriented Classification Step 3 –Begin creating class hierarchy –Training segment selection and standard nearest neighbor –Nearest Neighbor Classification of each level Step 3 –Begin creating class hierarchy –Training segment selection and standard nearest neighbor –Nearest Neighbor Classification of each level
Segmentation and Object- oriented Classification Step 3 –Begin creating class hierarchy –Training segment selection and standard nearest neighbor –Nearest Neighbor Classification of each level Step 3 –Begin creating class hierarchy –Training segment selection and standard nearest neighbor –Nearest Neighbor Classification of each level
Segmentation and Object- oriented Classification Step 3 –Begin creating class hierarchy –Training segment selection and standard nearest neighbor –Nearest Neighbor Classification of each level Step 3 –Begin creating class hierarchy –Training segment selection and standard nearest neighbor –Nearest Neighbor Classification of each level
Utilize the layer 7 mask The classified layer must be class 3 Utilize spatial attributes Turf and Grass must border residential or other turf and grass Turf and Grass must have residential in level 3 Utilize other levels Water in Level 1 is based solely on the existence of water in Level 3 Water in Level 3 is based on the summer red band (layer 4) and the standard nearest neighbor samples Segmentation and Object- oriented Classification Step 4 –Adding knowledge to each eCognition project –Refinement and final classification with class-related features Step 4 –Adding knowledge to each eCognition project –Refinement and final classification with class-related features
Site 1, Date 1Site 1, Date 2Site 2, Date 1Site 2, Date 2 Segmentation and Object- oriented Classification Step 4 –Adding knowledge to each eCognition project –Refinement and final classification with class-related features Step 4 –Adding knowledge to each eCognition project –Refinement and final classification with class-related features
Segmentation and Object- oriented Classification Step 5 –Use the knowledge- based classifier in ERDAS Imagine to do a post- classification change detection –Final change classifications for Site 1 and Site 2 Step 5 –Use the knowledge- based classifier in ERDAS Imagine to do a post- classification change detection –Final change classifications for Site 1 and Site 2
Segmentation and Object- oriented Classification Step 5 –Use the knowledge- based classifier in ERDAS Imagine to do a post- classification change detection –Final change classifications for Site 1 and Site 2 Step 5 –Use the knowledge- based classifier in ERDAS Imagine to do a post- classification change detection –Final change classifications for Site 1 and Site 2 Site 1Site 2
ResultsResults –Post Classification Analysis –Cross-correlation Analysis –Neural Networks –Segmentation & Object-oriented Classification
Post Classification Analysis Site 1 Site 2 UrbanAgricultureForestWaterBarrenAgr to Urban Forest to Urban Barren to Urban
Cross-correlation Analysis Site 1Site 2 UrbanAgricultureForestWaterBarrenAgr to Urban Forest to Urban Barren to Urban
Neural Networks Site 1Site 2 UrbanAgricultureForestWaterBarrenAgr to Urban Forest to Urban Barren to Urban
Segmentation and Object-oriented Classification Site 1Site 2 UrbanAgricultureForestWaterBarrenAgr to Urban Forest to Urban Barren to Urban
September 3, 1989 September 23, 1999 Post-classification Change Detection Post-classification Change Detection Agriculture To Urban Forest To Urban Barren To Urban
September 3, 1989 September 23, 1999 Cross-Correlation Change Detection Cross-Correlation Change Detection Agriculture To Urban Forest To Urban Barren To Urban
September 3, 1989 September 23, 1999 Neural Network Change Detection Neural Network Change Detection Agriculture To Urban Forest To Urban Barren To Urban
September 3, 1989 September 23, 1999 Object-oriented Change Detection Object-oriented Change Detection Agriculture To Urban Forest To Urban Barren To Urban
ConclusionsConclusions The results of this research reveal that there is merit to each of the several land use change detection methods studied, but that there appears to be no single best way in which to perform change analysis The most significant conclusion of this study is that much research remains to be done to improve upon the results of land use and land cover change detection
RecommendationsRecommendations These investigators firmly believe that an approach based on image-segmentation and rule-based classification is potentially such an improved methodology, and accordingly intend on pursuing the avenues of neural network and object-oriented classification change detection, perhaps in an integrated approach.
AcknowledgementAcknowledgement National Aeronautics and Space Administration Grant NAG /NRA-98-OES-08 RESAC- NAUTILUS, Better Land Use Planning for the Urbanizing Northeast: Creating a Network of Value- Added Geospatial Information, Tools, and Education for Land Use Decision Makers. Northeast Applications of Useable Technology In Land planning for Urban Sprawl
This presentation is available at resac.uconn.edu
A Comparison of Land Use and Land Cover Change Detection Methods Daniel L. Civco, James D. Hurd, Emily H. Wilson, Mingjun Song, Zhenkui Zhang C enter for L and use E ducation A nd R esearch Department of Natural Resources Management & Engineering The University of Connecticut U-4087, Room 308, 1376 Storrs Road Storrs, CT