Hay_Topic 10 - Raster analysis_v1 (1).pdf

Geog 380: Geospatial Communication Source: https://earth.nullschool.net Topic 10: Raster Analysis GEOG 380 - Topic 10 © Geoffrey Hay (2023) 1 Learning outcomes By the end of this lecture topic, a successful student will be able to: 1. Explain image enhancement 2. Explain the difference between...

Geog 380: Geospatial Communication Source: https://earth.nullschool.net Topic 10: Raster Analysis GEOG 380 - Topic 10 © Geoffrey Hay (2023) 1 Learning outcomes By the end of this lecture topic, a successful student will be able to: 1. Explain image enhancement 2. Explain the difference between point, neighbourhood, and zonal operations with raster data 3. Explain the use of neighbourhood (convolution) operators for analyzing raster data GEOG 380 - Topic 10 2 GIS Analysis: Raster § Raster have some qualities that impact our abilities to process them in a GIS environment § Pixels are a constant size within layers § Typically, there is only one attribute (i.e., DN values) per entity in any given layer § Entities and attributes exist regularly and continuously (i.e., wall-to-wall) § Raster analyses tend to be FASTER than vector analysis, and the basic functions SIMPLER § However, strung together into algorithms, raster processing can be very complex, and many types of transformations can ONLY occur in a raster environment GEOG 380 - Topic 10 3 Raster Data Operations § Many vector transformation operations do not apply in the raster environment. § Global raster operations § Done on entire image - enhancements § Three basic types of raster operations: 1. “Point” or local operations 2. Neighborhood or convolution operations 3. Zonal operations GEOG 380 - Topic 10 4 Image Enhancement § Purpose: improve the visual appearance and interpretability of an image § Contrast manipulation § Grey-level thresholding § Density slicing § Contrast stretching GEOG 380 - Topic 10 5 Image Thresholding § Used to segment an input image into two classes § One class for values below a specified DN § One class for values above a specified DN § Often used to prepare a binary image to separate (i.e., mask) spectrally-distinct features for further analysis GEOG 380 - Topic 10 6 Image Thresholding Original Image - NIR Threshold (DN = 47) Original Image: False Color – veg is red GEOG 380 - Topic 10 7 Density (Level) Slicing § DNs along the X-axis of the histogram are segmented into analystdefined intervals (slices) § Similar to thresholding, except it involves numerous classes § Simplistic approach for image classification GEOG 380 - Topic 10 8 Density Slicing Both density slicing and thresholding are typically applied to just a single band GEOG 380 - Topic 10 9 Contrast Enhancement § Original DN values rarely extend over the entire output range of a display device § Contrast enhancement techniques enhance or ‘stretch’ original data to accentuate contrast/interpretability of image § Depending on technique used, data integrity may be affected GEOG 380 - Topic 10 10 Contrast Stretching GEOG 380 - Topic 10 11 http://www.nrcan.gc.ca/earth-sciences/geomatics/satellite-imagery-air-photos/satelliteimagery-products/educational-resources/9389 Contrast Enhancement Techniques Linear contrast stretch GEOG 380 - Topic 10 Histogram equalization 12 Histograms and Lookup Tables Exponential Stretch 255 255 DN (output) 0 255 255 DN (input) GEOG 380 - Topic 10 Number of Pixels Number of Pixels Number of Pixels DN (output) 255 0 0 Logarithmic Stretch DN (output) Linear Stretch DN (input) 255 DN (input) 13 Point Operations Old Values New Values 1 0 2 0 3 0 4 0 5 1 6 0 7 0 8 0 9 0 GEOG 380 - Topic 10 § Operations that calculate new values for a raster grid on a cell-by-cell basis § Simplest form: reclassification of existing raster values § Actually, a very powerful technique that operates similar to a query (vector), and can be combined with other point operations to accomplish many things 14 Point Operations: Raster Calculator § Map algebra involves mathematical or boolean operations to be calculated amongst various raster layers on a pixelby-pixel basis § Exceptionally powerful and flexible § Raster 1 Raster 2 2 6 5 4 6 7 5 3 2 4 5 6 3 8 4 3 GEOG 380 - Topic 10 Newraster 0 0 0 0 * 0 0 0 1 0 0 1 1 0 1 1 1 e.g., Raster1 * Raster 2 = Newraster 0 0 0 0 = 0 0 0 3 0 0 5 6 0 8 4 3 15 Convolution or Neighbourhood Operation http://www.nrcan.gc.ca/earth-sciences/geomatics/satellite-imagery-air-photos/satellite-imageryproducts/educational-resources/9389 GEOG 380 - Topic 10 § Convolution involves the passing of a moving window over an image and the creation of a new layer § Each pixel in the new layer is a function of the original pixel values underlying the convolution window and the coefficients of the convolution window as specified by the user § Also known as ‘moving window’, ‘local’ or “filtering” operations 16 Convolution Operation Pixel Values Averaged or Removed From Raster GEOG 380 - Topic 10 17 Convolution: Edge Effects Edge pixels averaged or removed from the raster GEOG 380 - Topic 10 18 Convolution Coefficients § The moving window (kernel) is typically a square matrix of convolution coefficients, commonly 3x3, 5x5, or 7x7 pixels in size ⨂ Convolution Variable GEOG 380 - Topic 10 19 ‘High-Pass’ and ‘Low Pass’ Filters § Low-pass filters remove the high frequency information (noise) and show general trends in the data § High-pass filters remove the low frequency information and show the high frequency GEOG 380 - Topic 10 20 Mean or Average Filter (Low Pass) 89 92 95 89 91 107 142 166 92 106 114 120 171 177 179 ⨂ Mean Filter 90 94 105 118 130 93 105 122 140 153 107 128 150 165 174 1/9 1/9 1/9 122 175 187 179 179 1/9 1/9 1/9 131 158 174 180 180 177 192 186 181 178 1/9 1/9 1/9 159 180 185 183 183 Original Raster GEOG 380 - Topic 10 Convolved Raster 21 Laplacian Filter (Edge Detector) 89 92 95 89 91 107 142 166 92 106 114 120 171 177 179 122 175 187 179 179 177 192 186 181 178 Original Raster GEOG 380 - Topic 10 ⨂ Laplacian Filter -1 -1 -1 -1 8 -1 -1 -1 -1 0 0 0 0 0 0 0 0 21 115 0 0 0 190 106 153 115 160 111 5 49 0 0 0 0 Convolved Raster 22 Convolution Applications § Convolution operations are widely applied in a variety of raster applications § Remote sensing texture analysis § DEM processing § Spatial filtering High frequency GEOG 380 - Topic 10 Low frequency 23 Zonal Operations Raster of values Raster of zones § § Resulting raster GEOG 380 - Topic 10 Similar to neighborhood operations, but the “window size” is variable, and defined by a zonal layer (grey tones) Operations available in a zonal context are similar to those used in neighborhood operations § In this example, calculation of mean values within each zone 24 Hypsometric tinting with hillshade overlay Hypsometric tinting (also called layer tinting, elevation tinting, elevation coloring or hypsometric coloring) is used to enhance elevation zones so map readers can better see differences in relief. Classed values GEOG 380 - Topic 10 Colour ramp https://www.esri.com/arcgis-blog/products/product/imagery/hypsometric-tinting/ 25 Unidirectional vs multidirectional hillshades https://pro.arcgis.com/en/pro-app/latest/help/analysis/raster-functions/hillshade-function.htm For more info see https://www.youtube.com/watch?v=c7nCuDbObcM GEOG 380 - Topic 10 26