Study Guide
Source Materials
Ma_2024_RS8-UnsupervisedClassification.ppt
Ma_2024_RS8-UnsupervisedClassification.ppt
Download original
Re-extract
18 slides.
Left = our PNG render of the original slide. Right = our extracted text (edit to correct).
slide-01
Image classification A process of assigning pixels in an image to one of a number of classes. As a result of image classification, a thematic map is generated. 1
slide-02
Image classification and thematic mapping Unclassified Classified L ege ndo fle velI 1Me ad ow 2Law n 3Oa k 4Res id 5Dryg rass 6Sw a m p 7Lake 8Salte v ap 9Co m /indu 10Tra n sp N W E S 2
slide-03
Classification process A classification scheme Classification methods Accuracy assessment 3
slide-04
Classification process A classification scheme Classification methods Accuracy assessment 4
slide-05
Classification schemes Classes are simplification and generalization of the real world. Defining a classification scheme depends on the information extraction objectives. You could design a broad type which is abstractive or specific type which is concrete. A hierarchical classification system is usually used. e.g. Anderson (1976) land-use and land-cover classification system for use with RS data 1. level has 9 classes: urban or build-up, agriculture, forest, water, 2. level has 37 subclasses 5
slide-06
Anderson’s Land use and land cover classification system Level I Level II 1 Urban or Built-up Land 11 Residential 12 Commercial and Services 13 Industrial 14 Transportation, Communications, and Utilities 15 Industrial and Commercial Complexes 16 Mixed Urban or Built-up Land 17 Other Urban or Built-up Land 2 Agricultural Land 21 Cropland and Pasture 22 Orchards, Groves, Vineyards, Nurseries, and Ornamental Horticultural Areas 23 Confined Feeding Operations 24 Other Agricultural Land 3 Rangeland 31 Herbaceous Rangeland 32 Shrub and Brush Rangeland 33 Mixed Rangeland 4 Forest Land 41 Deciduous Forest Land 42 Evergreen Forest Land 43 Mixed Forest Land 6
slide-07
Anderson’s LULC classification system Level I Level II 5 Water 51 Streams and Canals 52 Lakes 53 Reservoirs 54 Bays and Estuaries 6 Wetland 61 Forested Wetland 62 Nonforested Wetland 7 Barren Land 71 Dry Salt Flats. 72 Beaches 73 Sandy Areas other than Beaches 74 Bare Exposed Rock 75 Strip Mines Quarries, and Gravel Pits 76 Transitional Areas 77 Mixed Barren Land 8 Tundra 81 Shrub and Brush Tundra 82 Herbaceous Tundra 83 Bare Ground Tundra 84 Wet Tundra 85 Mixed Tundra 9 Perennial Snow or Ice 91 Perennial Snowfields 7 92 Glaciers
slide-08
Classification process A classification scheme Classification methods Accuracy assessment 8
slide-09
Multispectral classification Make use of spectral response patterns of ground objects. Per-pixel or point classifiers Form the heart of the application of remote sensing in discrimination of land-cover types and conditions. 9
slide-10
Spectral reflectance curves Different types of earth surface features have their own distinctive electromagnetic reflectance properties. 10
slide-11
Multispectral Satellite Imagery Classification Supervised Classification Unsupervised Classification Maximum Likelihood ISODATA classification a priori knowledge is More computer- required automated more controlled by users 11
slide-12
Unsupervised classification (clustering) Only some parameters are required to specify from the user to begin this process. Then the computer uses these parameters to uncover statistical patterns (similar spectral characteristics) that are inherent in the data. Spectral classes do not necessarily correspond to any meaningful characteristics of ground objects. After classification, the users must attach the actual meaning to the resulting classes. 12
slide-13
(ISODATA) Iterative Self-Organizing Data Analysis Technique Uses minimum spectral distance to assign a cluster for each candidate pixel. Is iterative in that it repeatedly performs an entire classification and recalculates statistics. Figure 6-12 Minimum spectral distance 13 (Source: ERDAS Field Guide 2002 p.232)
slide-14
Three parameters for ISODATA Three parameters must be specified: N – the maximum number of clusters to be considered. Since each cluster is the basis for a class, this number becomes the maximum number of classes to be formed; T – a convergence threshold, which is the maximum percentage of the pixels whose class values are allowed to be unchanged between iterations; M – the maximum number of iterations to be performed. 14
slide-15
T – a convergence threshold After each iteration, the normalized percentage of pixels whose assignments are unchanged since the last iteration is calculated. # of pixels whose class value has Pixel Class value of Class value of Change of been changed = 3 previous iteration current iteration class value # of pixels whose class value has not been changed = 7 1 1 1 0 T (unchanged percentage of pixel 2 2 2 0 reassignments) 3 3 3 0 = # of unchanged pixels 4 4 4 0 5 4 3 1 total # of pixels 6 3 4 1 = 7/10 = 70% 7 2 1 1 8 2 2 0 9 3 3 0 10 4 4 0 15
slide-16
ISODATA clustering procedure (1) Begins by determining N arbitrary cluster means. The spectral distance between the candidate pixel and each cluster mean is calculated. The pixel is assigned to the cluster whose mean is the closest. After each iteration, the means for each cluster are recalculated, based on the actual spectral locations of the pixels in the clusters, causing them to shift in feature space. Then these new “means” are used for defining clusters in the next iteration. 16
slide-17
ISODATA clustering procedure (2) The entire process is repeated: each candidate pixel is compared to the new cluster means, and assigned to the closest cluster. The process will terminate until either the convergence threshold T or the maximum number of iterations M is reached. 17
slide-18
ISODATA clustering procedure (1) (2) (3) (1) Five arbitrary cluster means in two-dimensional spectral space (2) ISODATA first pass (3) ISODATA second pass (Source ERDAS Field Guide, Figure 6-3, 6-4, and 6-5, pp. 215-216) 18