Parameter Setting Workflows
There are two philosophies to operating hyperspectral sensors, and the differences revolve around how a user decides to set the frame period/frame rate, or the frequency at which data frames are collected. We have the science-first approach and the operations-first approach. Both methods have validity, and which one you choose will greatly depend on the conditions of your operation.
Sensor Parameters
The sensor's frame period is a measure of how fast each frame (row of pixels) should be captured. Because the unit is time (in milliseconds), a fast frame period will be a lower value than a slow frame period. Frame period can be derived from the following equation. For most systems, especially those utilized solid state drives (SSD) for storage, frame readout time is negligible.
Exposure Time is a measure of the integration time of light for each captured frame, which determines how saturated the frames will be. To maximize radiometric quality, exposure should be set so the spectral regions of interest are sufficiently saturated.
Differences in Practice
In the science-first approach, users will set their exposure (and gain mode) based on the available solar radiation, then set their frame period as close to the exposure as possible (as long as it is slower than than the maximum possible frame rate of the sensor). Then update flight plan speed to cooperate with the frame period.
In the operations-first approach, users will plan and fix their flight speed to meet their operational requirements. Because flight speed is fixed, frame period will also have a fixed maximum value (aka minimum capture speed). Then, without adjusting frame period, users then set exposure time as high as needed to fully saturate the sensor, potentially modifying gain mode as well.
Workflow
Science-First
Set sensor exposure and gain mode to the point where desired saturation is achieved
Set frame rate/frame period as close to necessary exposure as possible
Set flight speed to match frame period (taking oversample rates into account)
Operation-First
Set flight parameters as necessary (given all sensor constraints) and leave fixed
Set frame rate/frame period to match planned flight speed, leaving fixed
Set exposure until desired saturation is achieved
If desired saturation is not achievable, operator must decide between adjusting flight plan, increasing gain level, or leaving data potentially underexposed
Considerations
Science-First
When atmospheric conditions are cooperative (peak solar noon, no clouds), the necessary exposure time tends to be quite low. This enables the aircraft to fly at relatively quick speeds.
Fast speeds can result in pushing sensor frame rate close to its limits.
Fast speeds also have negative impacts on aircraft attitude (pitch) and can have negative consequences on orthorectification alignment.
Requires flight plans to be adjusted in the field. At the very least, speed needs to be updated based on the necessary exposure/frame period.
Users will not know how long their flight will be, nor whether the area of interest can be surveyed in a single flight, until after setting up the hyperspectral sensor. This may require splitting flight plans in the field, building new capture polygons, etc.
This method results in minimizing the gap between exposure time and capture time. Therefore, each spatial pixel more closely represents the underlying scenery.
When flying multi-modal systems like GRYFN Gobi, other integrated sensors may not be as agile in flight parameter changes while ensuring maximum data quality.
This method works best operationally when survey areas are small and conservative flight time estimates are not close to the aircraft's maximum flight time, and when frame cameras are not involved. This method excels from a data quality standpoint in unknown lighting conditions, and because each pixel captured will more accurately relate to the scenery spatially.
This method suffers from decreased efficiency given that operators must adjust flight plans in the field, especially if large changes are needed or battery sets are limited. This method also poses risks on multi-modal systems where other sensors are more reliant on fixed flight parameters for best processing results.
Operation-First
When operating hyperspectral sensors, operators typically will only want to fly in good atmospheric conditions, where minimal exposure is needed and therefore frame periods do not necessarily need to be adjusted.
Flight plans remain fixed. User can be more efficient with in-field setup. User knows exactly how long a flight will be, and can properly plan for multi-flight missions, have additional capture polygons prepared, etc.
This method does not minimize gaps between exposure and capture time. Therefore, each frame is artificially "stretched" to fit the full frame period.
Ex: 4ms exposure time, 6ms frame period time. 4ms of exposure is "stretched" to fit the 6ms frame time.
When flying multi-modal systems, users can more easily and reliably plan for flight parameters that work for all sensors at once.
This method works best when users may not have any additional flight time to extend flights should speed need to change for excess frame period, when multiple sensors are involved in concurrent capture, and when operational efficiency is highly important. This method ensures multiple sensors can have good data quality simultaneously with little to no user intervention.
This method suffers in poor lighting conditions where flight speed/frame period doesn't allow for sufficient saturation of DNs. Users need to decide whether to adjust their flight plan or undersaturate their data (not often a problem as hyperspectral sensors are mostly flown in bright and consistent sky conditions). This method also suffers in micro analysis where individual pixel spectra, rather than area averages, are of the highest importance.
Adjusting Exposure for Scenery
GRYFN has available a small 82% calibrated reflectance panel to aid users in setting hyperspectral exposure levels. Our typical recommendation for setting hyperspectral exposure levels is to adjust exposure of the hyperspectral sensor to 90-95% saturation while looking at the 82% exposure reference panel.
Our saturation/exposure recommendations using the 82% exposure reference panel are generalized, typically providing good results across a variety of applications. However, given the generalization, in certain environments this may lead to over or underexposure of important imagery. We urge users to set exposure to achieve sufficient saturation for their goals. This takes experimentation, and may mean you may need more or less exposure/saturation than we recommend depending on the environment, scenery, and what's important for your research.
If resulting saturation using our recommendation is too low or too high, you may have to adjust exposure outside of our recommendations. In cases where you are slightly overexposed on the 82% panel, some parts of the scenery may end up overexposed, but again, this may or may not have a negative impact on analysis.
Last updated