1. Image Processing & GIS Integration for Environmental Analysis School of Electrical & Electronic Engineering The Queen’s University of Belfast Paul Kelly

2. Objective Environmental Analysis Currently two extremes –Satellite image only low-resolution roof / surface canopy visible –Detailed individual plant samples Gap currently filled by manual subjective scene analysis Economic & Versatile method of surveying moderately-sized areas needed

3. Existing Research Agouris (’00) GIS integration with aerial photographs Broggi (’98) Perspective transformation, only for flat scenes McMenemy [QUB] (’01) DV Camera Calibration and use in photogrammetry Multi-disciplinary subject

4. Data Types & Processing Raster: pixel-level (colour) Vector: object-level Sources –Ground-level Images captured with Digital Video camera (raster) –GIS Digital Map Data (mainly vector) –Aerial Photographs or Satellite Images (raster)

5. Change Detection Preliminaries Comparison only possible between images of the same view of same scene These parameters must be equal: –Observer (camera) location –Centre of View –Camera Parameters (Field of View, focal length, CCD Array Dimensions) Perspective Transformation transforms an image so they appear equal

6. Methodology Outline 1. Initial Image Correction 2. Camera Location Measurement 3. GIS Data 4. Ground-level Image / GIS Processing 5. Aerial / Satellite Image Integration 6. Change Detection (Preliminary Results)

7. Initial Image Correction Colour Response Calibration –Dark Image –Response Curve calibration with Luxmeter Lens Barrel Distortion –Image of grid of known points –Best-fit transformation equation –Bi-linear Interpolation

8. Original Distorted Distortion Correction Original Distorted Corrected

9. Camera Location Measurement GPS Accelerometer Electronic Compass Sensor Fusion and processing results in Position (Easting, Northing, Elevation) and Orientation for GIS input

10. GIS Data Supplied by Ordnance Survey (NI): — 1:1250 or 1:2500 scale Vector map data — 50 metre resolution digital elevation data — 1:50000 scale contour lines — Spot heights (mainly on main roads and benchmarks) included in Vector maps Feature & Building Outlines etc. Elevation / 3rd Dimension data, to be interpolated into a Digital Elevation Model

11. Ground-Level Images Commercial Digital Video DV Camera Experimentally calibrated for –Field of View –Focal Length –CCD Array Dimensions Thus enabling –Photogrammetric Calculations –Generation of ‘Camera-eye Views’ in GIS or otherwise

12. GIS 3-D View 3-D Landscape View Field Boundaries overlaid on Image

13. Aerial / Satellite Images Require georeferencing in GIS— polynomial approximation may be used Images used for –Initial detection of where change has occurred—pointer to more detailed ground observation and analysis –Framework for referencing ground-level images to

14. Aerial / Satellite Images Russian satellite image from 1988 Blue lines: edge- detected from satellite image Red lines: GIS Building Outline Data

15. Change Detection Methods Illumination and Daylighting Compensation (Use calibration object) Vector / Vector Comparison –‘Rasterised Vector’ Image created by segmentation and edge detection –Objects displaced or changed shape? Raster / Raster Comparison –Surface features, e.g. colour changed?

16. Further GIS Use GIS 3-D View related pixel-by-pixel to original image Calculate Easting, Northing & Elevation for each pixel Use referenced pixel values to re-create the view from a different perspective Works well in areas with few objects above ground

17. 6 January 2002 24 April 2002 Further GIS Use Aerial Views With some illumination compensation Difference Image

18. Summary Calibrated Camera Images GIS Digital Map Data Innovative Image Alignment Methods Vector & Raster Change Detection GIS use significantly enhances Image Processing & Photogrammetry capabilities