This write-up intends to present birds eye view of fully automated and accurate mapping solutions based on ultra-light UAV imagery. We showcase interesting observations in the field of UAV mapping, the steps to analyze the accuracy of the automated processing on several datasets. The software used to process is one of the leading and evolving software in the UAV data processing domain.
The accuracy highly depends on the ground resolution (GSD) of the input imagery. When chosen appropriately this mapping solution can compete with traditional mapping solutions that capture fewer high-resolution images from airplanes and that rely on highly accurate orientation and positioning sensors on board. Due to the advancement of computing practices and processing prowess of computers and careful integration with recent computer vision techniques, the result is robust and fully automatic and can deal with inaccurate position and orientation information which are typically problematic with traditional techniques.
SBL’s geospatial team is one of the first in the region to process such images. Processing of UAV images has its own challenges. SBL used to receive post-processed UAV images along with IMU and GCPs as input. Aerial Triangulation is the first step performed. During this stage Ground Control Point (GCP) and Actual Check Point (ACP) reports has been generated. This is an iterative step till we get desired accuracy. The following will explain in brief some of the critical steps in the processing of UAV data.
1.The software examines for matching points by analyzing all images. The software used here an improved version of the binary descriptors, which are very powerful to match image points quickly and accurately.
2.Those matching points as well as estimated values of the image position and orientation provided by the UAV autopilot are used in a bundle block adjustment to reconstruct the exact position and orientation of the camera for every acquired image.
3.Based on this re-establishment the matching points are corroborated and their 3D coordinates calculated. The geo-reference system is WGS84, in this case, based on GPS measurements from the UAV autopilot during the flight.
4.Those 3D points are interpolated to form a triangulated irregular network in order to obtain a DEM. At this stage, construction of a dense 3D model increases the spatial resolution of the triangulated data.
5.This DEM is used to project every image pixel and to calculate the geo-referenced ortho-mosaic. The ortho image will be devoid of positional and terrain displacement inaccuracies.
One of the major application for which UAV images used are for agriculture. UAV images are ideal for small size farms. Plant counts such as corn counting will give an idea of yield from those plants. Plant health monitoring, differentiating species of agricultural farms/plants and plantation estimation are the major task performed for agriculture. Growth stages of the farms can also be monitored using ortho images acquired through UAV process. UAV image processing is also helpful for the site selection for solar farms.
In case of forestry, UAV images are very helpful in species identification. SBL’s interpreters have identified forest species and enabled the client to map forest land parcels. It is a tool to monitor de-forestation as well as afforestation. Golf courses are another field where UAV data sets are highly useful. Golf course features can be mapped with their actual heights through this process.
Mining industry is the most benefitted in the usage of UAV technological advancement. As most of the mines are spread over small areas, UAV data acquiring and processing is very cost effective. Along with other data processing, SBL has got the expertise in generating contours and mining related features to very minute levels of detail.
SBL - UAV Image Processing
- Aerial Triangulation
- 3D Feature Extraction