Collecting and Processing Sub-Inch Per-Pixel Resolution Aerial Imagery for Geospecific Real-Time Terrain
MetaVR has established a low-cost aircraft data collection and processing workflow which yields real-time 3D terrain. The terrain is comprised of geographically specific 2 cm per-pixel resolution imagery. Using MetaVR’s Terrain Tools for Esri® ArcGIS® to combine high-resolution imagery of an area of interest with accurate satellite elevation data or other digital elevation models (DEMs) results in a realistic geospecific synthetic environment rendered in MetaVR’s Virtual Reality Scene Generator™ (VRSG™).
The workflow includes content creation using Autodesk 3ds Max and other modeling tools to create custom 3D models of any or all of the surrounding area elements, based upon aerial and/or ground-level photographs. Content can be relevant and specific to various kinds of simulation training, such as military mission exercises, law enforcement, or emergency/disaster response.
The MetaVRC™, MetaVR's small UAV for imagery data collection, is a low profile, lightweight, and portable plane, built for autonomous aerial surveying. The purpose of the MetaVRC is to take high-resolution still-frame images that are then orthorectified and can be used for building 3D terrain for rendering in VRSG.
The aircraft, built by Swift Radioplanes (SRP), is designed to fly at low altitudes in order to capture images at sub-inch resolution.
The images on this webpage illustrate the process of collecting and processing the imagery and then using the imagery for building terrain and 3D content. The examples come from recent deliveries to customers and a sample test dataset built from imagery of an arid rural area that SRP provided to MetaVR.
As part of a simulation technology upgrade to the training dome and desktop systems in 2015, MetaVR delivered to the Naval Aviation Warfighting Development Center (NAWDC) at Naval Air Station (NAS) Fallon, NV, high-resolution geospecific terrain of two target ranges on the Fallon Range Training Complex (FRTC) used for field training in the facility's JTAC qualification course. The 2 cm per pixel imagery of the B-17 and B-19 ranges was captured by the MetaVRC over the course of 9 days. The remote-controlled aircraft was flown by SRPs in military controlled airspace. With the aircraft’s 24 MP and 36 MP cameras and NED 10 meter elevation data, MetaVR compiled full-resolution (2 cm) terrain tiles of the ranges with MetaVR Terrain Tools for Esri ArcGIS. The total area of coverage of this terrain data set is 65.81 sq km.
At 2 cm resolution, details such as helicopter landing areas, vehicle targets, and small craters left from exploded ordnance are visible on the terrain and bullet holes are visible on targets. (Underlying the high-resolution 3D terrain of the two ranges is MetaVR's CONUS ++ terrain, which was built with 1 meter per-pixel terrain imagery and DTED-1 elevation data.) Click here to see a real-time video recorded in VRSG featuring a flyover scenario of the Fallon Range terrain.
The following VRSG screenshot shows the main road of terrain of an early imagery-collection testing area built with 2.5 cm imagery. The imagery of this road was captured with a small rotary-wing UAV at a higher, 5 mm, resolution. As a validation of the technology, the screenshot shows how well the 5 mm resolution road blends into the rest of the 2.5 cm terrain at the shoulder of the road.
Aircraft: Lynx fixed-wing SUAS, developed by SRP
Launching the aircraft
The aircraft is hand-launched, with two people typically handling the take-off. One person physically throws the plane, the other handles the GCS.
The motor and propeller combination of the aircraft, its light weight, and large wing span make hand-launching easy. The MetaVRC produces enough static thrust to basically pull the aircraft out of the hands of the person who throws it.
Recovering the aircraft
The aircraft lands via a vertical deep-stall method, which enables operations and recovery in confined areas, such as a forest clearing. The aircraft, with only three moving parts, is extremely durable to withstand field abuse. Two people typically handle the recovery. One person physically recovers the plane, the other handles the GCS.
The same components that allow for tool-less assembly also allow the aircraft to come apart upon landing, thus avoiding damage from landing impact. (The aircraft is fitted with durable crash pads beneath the fuselage to absorb the impact, and the entire aircraft is constructed of tough Kevlar and foam.)
Maximum payload weight: 700 grams
Preparing to collect the raw imagery data
Before launching the MetaVRC to capture imagery data of a given area, control points (physically marked waypoints, visible to the plane’s camera) are placed at various locations on the ground around the area of interest. These control points create a set of geographically referenced points in the aerial imagery -- denoting the exact ground location of an area geographically when the imagery photos are taken. These ground coordinates increase the accuracy of the aerial imagery or can be used as checkpoints to verify the accuracy.
Setting up the ground control points varies based on size, terrain, physical access, transportation, type of survey, and so on.
Data collection for an area up to 8 sq km takes approximately 3 hours, yielding about 4000 raw images. Weather (wind), desired image overlap, and desired resolution (altitude) can cause this duration to increase or decrease. The best time to fly the aircraft for imagery collection is solar noon. With a low winter sun that timeframe is approximately 11 AM to 3 PM.
Ground control station and platform software
SRP's GCS software (Swift GCS) for the aircraft runs on Windows, Linux, or MacOS, and is typically run on a tablet for field use. The on-board Pixhawk Autopilot autonomously controls the aircraft in flight.
The aircraft, which has only three moving parts, is hand-assembled, without tools. It takes approximately 10 minutes to set up and launch the aircraft after removing its components from the two hard-back Pelican 1740 and 1500 transport cases and initializing the GCS software on the Windows notebook computer or tablet.
PROCESSING THE IMAGERY
Raw aerial images require geometric correction, or orthorectification, with GIS tools to remove distortions, enforce a uniform scale of distances, and improve the positional accuracy (using the ground control points) so that they depict an accurate representation and correspond to real-world coordinate systems.
MetaVR’s aerial imagery provides high-resolution geo-referenced orthomosiacs. During orthorectification, the high-resolution images collected by the MetaVRC are registered with real-world coordinates and overlaid onto DEMs to simulate the 3D terrain environment.
The orthorectification process of an imagery data collection can take a couple of days to a week, assuming the orthorectification results are satisfactory. During this process the images are mosaiced (assembled) into a uniform tiling scheme. After the imagery has been orthorectified, it is ready to be used to build 3D terrain.
Building 3D terrain with the orthomosiaced imagery
The orthomosaiced images, together with elevation data of the photographed region become source data inputs for MetaVR Terrain Tools to compile the terrain tiles. The resulting terrain tiles can be rendered in VRSG at 60 Hz.
With imagery at this resolution, you can create a physics-based IR profile of the terrain with a high degree of realism. Using the IR Setup utility that is delivered with VRSG, you would describe the sensor’s spectral response within its waveband of interest, train the material classifier, and specify the environmental characteristics that influence the appearance of the IR scene. The material classification process automatically generates material attribution from visual spectrum colors. This means that the higher the resolution of the visual database, the more accurate the IR profile.
CREATING 3D content from ground-level photographs
Aerial imagery can be correlated with ground-level photography in order to make 3D models of particular features of interest. Perspectiveless photographs are taken with at least a 16-megapixel camera of the feature of interest, ideally in overcast conditions. One set of photographs are taken to give an overall reference of the subject and then key reference features of the building such as doors, windows and so on are photographed by zooming in to focus on the key feature. Again, perspectiveless photographs are taken such that they can be used for creating textures on the 3D model.
While creating 3D building models from aerial photography or satellite imagery is common, such models are generally not useful for ground-level orientation as they lack resolution in both geometry and textures. Typically a heterogeneous terrain model is made from overhead imagery, 3D models are made from imagery for intermediate resolution, and finally 3D content is made from ground-level photographs to provide the highest level of detail.
Ordering aerial imagery or terrain
Aerial imagery collected by the MetaVRC can be ordered directly from MetaVR by specifying the area of interest (AOI), such as an airfield, or a training area.
An aerial imagery order consists of:
Aerial imagery orders must be for an area of a minimum of 20 sq. km. Access to the area of interest for aerial photography must be in accordance with FAA regulations. Customers will be responsible for obtaining the necessary authorization and certified access to operate a restricted airspace of interest for aerial photography.
Customers have unlimited unrestricted use rights to the imagery. MetaVR retains the right to use the collected data and resulting terrain tiles and to redistribute the terrain tiles (typically as part of a larger set of terrain tiles of a region).
To request a quote for aerial imagery or terrain product, contact MetaVR.