A standard visual sensor collects red, green and blue wavelengths of light. Multispectral sensors are able to collect these visible wavelengths as well as wavelengths that fall outside the visible spectrum. These include near-infrared radiation (NIR), short-wave infrared radiation (SWIR) and others.
“Multispectral sensors capture data on the reflection of light energy off objects in the environment. That data can be compared to other nearby objects to understand crucial difference between them,” shared Scott Hatcher, a Geospatial Scientist. “To go further, to universally compare reflectors, you must use calibrated sensors and process them differently. Whether to pursue this or not is dependant on the use case. Define your problem!”
Multispectral sensors contain between three to five spectral bands and fall into two categories: modified and multiband.
Modified sensors are created when a special filter is placed on a standard visual sensor. As a result, modified sensors collect three bands of light at once through the same lens. Filters can come in many different formats to display different combinations of spectral bands. The most common formats sacrifice one of the visual bands to record near-infrared information (NIR). For instance, an R-G-NIR filter sacrifices blue in order to collect near-infrared energy (~700 - 800 nm).
Multiband sensors are manufactured specifically for multispectral data collection. Each band is collected by a dedicated sensor so there is no need for multiple flights. Multiband sensors also enable you to mix different band combinations to meet your needs.
Multispectral sensors are the workhorses of drone-based advanced sensing. Their ability to capture data at exceptional spatial resolution—as well as determine reflectance in the near infrared—makes them an extremely versatile and effective sensor.Scott Hatcher, Geospatial Scientist
Uses of Multispectral Sensors
Multispectral sensors are instrumental in plant health and management. They can pinpoint nutrient deficiencies, identify pest damage, optimize fertilization and assess water quality.
Here are a few examples of industry use cases:
Review data related to environmental mitigation activities such as monitoring plant and tree health.
Identify instances of vegetation encroachment before damages occur.
Assess water quality following floods to prioritize relief efforts and better distribute emergency resources.
Perform more effective crop management with plot-level statistics on plant count, height, vigor, leaf area and canopy cover.
Review health of forestry and vegetation to determine which portion of land to clear for commercial projects.
Flight Planning and Data Collection
The amount of light reflected and absorbed by an object is called reflectance. When collecting data with multispectral sensors, reflectance panels should be used.
These panels are placed on the ground and reflect light at a consistent level. Before and after each flight, secure images while flying over the panel. This enables date, time and location data to be tied to lighting conditions and is necessary for calibration during data processing.
“When you collect data over very homogeneous areas such as mature corn crops, be sure you overlap imagery and fly at a consistent altitude and speed. This will prevent problems when images are stitched together during processing,” stated Matt Tompkins, Director of Flight Operations.
Lastly, when collecting data with a modified sensor, use consistent camera settings during each flight.
Data Management and Delivery
Before leaving the field, be sure to perform a quick review to ensure you captured all required data. When using a multiband sensor, this can be done with the specialized software created by the sensor manufacturer. It generally takes a few hours to process multispectral data and create an orthomosaic (stitched image).
Multispectral Sensor Comparison
|DJI Zenmuse X4S RGNIR||DJI Zenmuse X5S RGNIR||MicaSense RedEdge-M™|
|Number of Bands||3 (red, green, near-infrared)||3 (red, green, near-infrared)||5 (blue, green, red, red-edge, near-infrared)|
|Outputs||DNG, JPEG, DNG+JPEG||DNG, JPEG, DNG+JPEG||RAW TIFF, GEOTIFF|
|Compatible Drones||DJI Matrice 200 and 600, DJI Inspire 2||DJI Matrice 200 and 600, DJI Inspire 2||DJI Matrice 100 and 600|
Once the data has been stitched into an orthomosaic, further analysis can transform the bands in indices that are sensitive to vegetation health and stress, segment the imagery into objects of interest, or create three dimensional models of the area using photogrammetry. This requires using GIS software or remote sensing software. Once completed, these data products can be the start of a complex algorithmic processing chain that produces actionable information.