Topic 12: Raster Analysis Flashcards
Similarities and differences between raster and vector analysis: Example- vector vs raster overlay
Vector clip:
- construct new coincident vertices at intersection points in both data layers
- merge line segments and split polygons
- eliminate polygons outside clip parameters
Raster clip:
- determine the range of pixels rows and columns for clip region
- eliminate pixels that fall outside the clip extent
Considerations for GIS analysis in the raster environment:
- rasters use simple data structure (regular grid) that facilitates computationally efficient analysis
- typically, one cannot perform vector and raster simultaneously - one must often convert vectors to rasters first
- raster transformations are computationally quite different than for vector datasets
- data resolution and spatial reference system affects measurements and multi-raster analysis
Raster data transformation and analysis operations:
local operations
- local operations on raster data refer to cell-by-cell operations
- can be performed on a single or multiple raster images
- most raster processing falls into this category
Local operations on multiple rasters
- summary statistics, such as mean, minimum, maximum, or majority (mode) values are common local operations
- local operations also useful in activities like change detection analysis (eg., multi-temporal landscape change)
Raster data transformation & analysis operations:
- map algebra on raster datasets
- Map algebra is the process of constructing simple numeric models based on overlaying raster layers and performing mathematical operations on the layers
- eg., detecting landscape change by subtracting one raster grid (recent) from another (historical)
Map algebra examples
Arithmetic
- Arithmetic multiplication
- can combine operators with mathematical transformations, such as the square root tool
Boolean
- Boolean And operator (Examines cells for positive vs zero numbers, produces binary maps)
Relational
- Greater than or equal to operator
- also produces a binary map
Raster data transformation and analysis operations
- neighbourhood
- spatial filtering (convolution)
- Neighbourhood operations on raster data refer to data computations at a focal cell based on the values of neighbouring cells
- involves a window of pixels that moves across all cells of the raster to perform the same computation over and over for each location (referred to as moving window analysis, spatial filtering, or convolution)
Low- pass vs high-pass filter
- Low-pass filter the high frequency information (noise) to show general trends (smoothing operation)
- High-pass filters remove the low frequency information and show the high frequency (sharpening)
Applications of neighbourhood operation
- remote sensing texture analysis
- DEM processing
- spatial filtering
Bigger the size of the window, the smoother the output raster?
true or false?
True
What is a block neighbourhood?
spatial filtering (convolution)
- instead of having one focal cell, the window moves so that the neighbourhoods are adjacent to each other
What happens to the edge pixels in a convolution operation?
They are averaged or removed from the raster
What does the mean filter out?
- It is considered a low pass filter, so it smooths out to show general trends
eg., 3x3 window would have a coefficient of 1/9
What is a Laplacian filter?
- It is an edge detector, gives bright lines
- has a coefficient of 4 or 8 in the center and a cross of -1
Raster data transformation and analysis operations:
- zonal raster operations
- zonal operations or “zonal statistics” refers to raster analysis restricted to regions or zones within the map
- similar to block neighbourhood, but the “window size” is not square or regular, but rather polygon shape
- zonal operations are sometimes very sensitive to where one defines the boundaries (MAUP)
