Unclassified Point Cloud

The IMU and post processed airborne GPS logs were used to generate the LiDAR point cloud from the waveform instrument data. Raw LiDAR strips were levelled to establish internal consistency, merged and 1km x 1km tiles in LAS v1.2 format were created. An automatic classification algorithm was applied in TerraScan

software to produce an initial classification of ground (2) and unclassified (1). High and low point noise were The ground classification was improved manually by visually scanning the ground surface and reassigning points from ground to unclassified to remove spikes and by assigning unclassified points to ground

where the ground surface lacked sufficient detail to describe the terrain (ie large TIN triangles).

Level 3 Classified LiDAR Point Data The IMU and post processed airborne GPS logs were used to generate the LiDAR point cloud from the waveform instrument data. Raw LiDAR strips were levelled to establish internal consistency, merged and 1km x 1km tiles in LAS v1.2 format were created. An automatic classification algorithm was applied in TerraScan software to produce an initial classification of ground (2) and unclassified (1).

High and low noise points were automatically classified. In accordance with Appendix B of the Statement of Requirement, in areas of overlapping swaths, points from the swath with the lowest off-nadir angle were classified to class 12 (overlap).

The ground classification was improved manually by visually scanning the ground surface and reassigning points from ground to unclassified to remove spikes and by assigning unclassified points to ground where the ground surface lacked sufficient detail to describe the terrain (ie large TIN triangles).

On completion of the ground classification automatic algorithms were used to classify unclassified points to low vegetation (3), medium vegetation (4), high vegetation (5) and buildings (10).

The workflow and quality assurance processes were designed to achieve the Level 3 requirement for 'removal of significant anomalies which remain in the ground class (2) and achieve a ground point misclassification rate of 2% or less. The classification accuracy was not measured.

LiDAR Intensity Image

Intensity images with a ground pixel spacing of one metre were generated from the LiDAR point cloud using an RPS algorithm that calculates the intensity image as the mean intensity of the first return points or last return points falling on each pixel. Both first and last return images were produced. The intensity images were converted to ECW using ERMapper software and also exported as 1km x 1km tiles in GeoTIFF

format.

Digital Elevation Model (DEM)

A DEM with a ground pixel spacing of one metre was generated from the LiDAR ground points. The method of deriving the height for each grid point is by triangulation between the three nearest LiDAR ground points. Hydro flattening was applied using waterlines (break lines) that were manually digitised using the LiDAR

and ortho photography as reference for areas of non-tidal water greater than 625 square metres and watercourses nominally greater than 30 metres in width.

The DEM was exported to ESRI Binary grid as 1km by 1km tiles in accordance with ICSM specifications (December 2012).

The DEM accuracy will be limited by the spatial accuracy of the LiDAR point data and will contain some additional error due to interpolation, particularly in areas of dense vegetation where ground points are sparse. The may also be a small amount of error due to ground point misclassification.