Flying Colours: Esper Demonstrates Aerial Hyperspectral Imagery
While we build our imaging systems to handle the harsh conditions faced in orbit, testing the quality of data that it captures and the processing we have to do to make a data product is much less straightforward than testing mechanical and electrical systems by shaking and baking them. We’ve decided to test an earlier version of our OTR (Over The Rainbow) imagers on a drone-based platform to accomplish this significant test.
The aerial test consists of a few caveats. Since we’re much lower than where we would fly when on a satellite, we have to come closer to our resolution estimates in space. This demands a change to our imaging optics, wherein in this test, we pick up an off-the-shelf Canon lens to work with our spectrometer. This still gives us a much higher resolution than what we expect. Hence we plan to reduce this resolution artificially. To test our data processing within space, we need to provide our systems with a similar data product. This is why this reduction in resolution is essential to validate our technology.
The aerial test primarily tests our capture and processing workflow but also, most importantly, tests our spectral accuracy in the field, out of the lab, capturing a landmass that we would typically see through our space platforms.
We went out to Kangaroo Ground in northeast Victoria to fly over a vineyard property supported by the guys at 3FB Aerworx and their drone capabilities.
Results
We’re delighted to say that we’ve achieved operational success. We’ve been able to verify the functioning of our spectrometer's electrical and software systems. This included capturing multiple exposures while being modulated with onboard telemetry data. This is the same way our computing systems will capture and process data in space, albeit with slightly different data. We were able to replicate this over five runs of the drone over the vineyard.
But how good is the actual imagery?
Below is the raw capture as seen by our imager. The ‘smudge’ in the middle of the image is where the spectral information is stored, with the left ‘bar of light’ being the zeroth-order diffraction. Unfortunately, our zeroth order is too overexposed for us to perform any stitching functions. While we might have good spectral data, we cannot create a full raster for the entire landmass.
That being said, we can confirm we have captured accurate spectral data. The below curves show the spectral response captured from the vineyard crops versus what has been recorded in literature with other COTS imagers and other satellite platforms. This induces confidence in our spectrometer design and our data processing systems. However, we are still somewhat limited by the optics we are using for our main imaging systems (the Canon lens), which cuts off reflectances in the infrared bands and misses out on many vital bands. We aim to rectify this in our future V3 design, which we’re looking to fly in the coming months.
