Let’s talk about calibration


Share | 09/01/2020

Table of

The value of radiometric calibration can be a bit difficult to understand,  so we’ve put together this article in an attempt to add clarity to the topic. Hopefully, after reading this you will understand conceptually how changes in day-to-day light conditions can impact the accuracy of multispectral data and what tools are available to improve radiometric accuracy.

What is Radiometric Calibration and why is it important?

Although we generally visualize multispectral imagery as colorful indices or composites, the actual images that come out of the camera are grayscale and are essentially matrices of digital numbers. Below is an example of the grayscale images of raw data captured with the MicaSense series RedEdge-MX:

The images are ordered from left to right as blue, green, red, red edge and near infrared.  You’ll notice that as the images move out of the color portion of the spectrum (blue, green, red) and into the near infrared (red edge and near infrared), the plants in the image get lighter.  Healthy plants mostly absorb visible light and by comparison, reflect quite a bit of red edge and near infrared light…that’s why you see darker plants in the left images and lighter plants in the right images.  We use this light and dark reflectance data at each band to help understand the physiological condition of a plant canopy…for example, if the red band shows extremely dark plants, those plants are absorbing a lot of red light and are photosynthetically active.  On the other hand, if the red band shows plants as lighter, they may be experiencing a stressor that’s interfering with photosynthesis.

Plant reflectance curve across different wavebands. Note that plant reflectance increases towards the near-infrared end of the spectrum, correlating with the plants in the raw images from the camera looking lighter as they get closer to the near-infrared band.

As mentioned earlier, you can think of images as matrices of digital numbers. Each pixel/cell within an image contains a digital number corresponding to the intensity of radiance within a certain wavelength. Each of the 5 images in this example has dimensions of 1280 x 960, meaning the total number of pixels in each image is 1,228,800.  Each of those pixels has a value assigned to it that when combined with all the other pixels. For example in the images above, that allows us to see things like buildings, driveways and grass in the images.  If we zoom into part of a tree in the NIR band image, we can see how the image is comprised of square pixels with different digital numbers:

Unfortunately, these pixel values are relative to the conditions in which the data was collected, and are not absolute. This is largely due to changes in light conditions (e.g. sunny vs overcast, sun at different points in the sky during the day, half of a field more heavily overcast than the other half, etc). If we are flying a crop over the duration of a season and are looking for subtle changes from flight to flight like nutrient deficiencies, early pest infestations or early disease identification, it becomes very important to capture as accurate of pixel values as possible and correct for any lighting changes that have impacted the data.  In order to detect actual changes in plant-canopy reflectance from multispectral images captured over two or more days (e.g. an early, mid and late season flight), it is necessary to perform a radiometric correction.

What do I need to use to calibrate my data?

Now that you know the ‘what’ and ‘why’, let’s discuss the ‘how.’  To get radiometric accuracy and repeatable results, you need a baseline reflectance measurement–a known reflectance reference point.  It can also help to know how the lighting conditions changed during the flight itself.

So how can we get a good quality baseline measurement and therefore accurate radiometric calibration? There are two standard methods, both which are common in remote sensing. The first and most historically used method is using something called a calibration panel. The panel has pre-measured reflectance values and therefore acts as a “control”. Taking a picture of the calibration panel allows you to assign the known reflectance values to the pixels of the panel and adjust the rest of the dataset accordingly.

MicaSense series Calibrated Reflectance Panel (CRP)

All MicaSense series camera kits come with the Calibrated Reflectance Panel (CRP) which we encourage our customers to use before and after each flight. If you take a panel picture not only before but also after each flight, you have two baseline measurements to work with and can also discern how the lighting conditions changed during the flight. Most processing softwares allow the user to upload their panel images and thus apply the radiometric correction. Some softwares allow users to upload both the before and after panel pictures as well. 

The second tool for calibration is called the Incident Light Sensor or Downwelling Light Sensor (DLS).  This is upward facing, mounted on top of the drone, and records data on lighting conditions throughout the flight, writing them into the metadata of each image captured which can later be used during image processing to fine-tune the radiometric correction done by the panel, enhancing its accuracy.

The process of radiometric calibration incorporates many key elements such as the position of the sun and sensor as well as irradiance data from light sensors and/or reflectance panels. The radiometric calibration process can take all this information and core sensor parameters such as camera gain and exposure to enable the process of converting digital numbers from raw multispectral imagery into sensor reflectance/irradiance and then into surface reflectance.


At this point, you know that radiometric calibration changes image pixel values to accurately represent the true reflectance of objects in an image.  Two main tools–a reflectance panel and incident light sensor–help us capture the information needed for radiometric calibration, and are necessary components of any multispectral imaging toolset. 

Because plant reflectance can be an indicator of health, stress, disease, differing varieties or species and more, accurate reflectance values are key to understanding plant physiology and comparing imagery from day to day or season to season.  Time-based analysis is not possible without accounting for lighting conditions, and is therefore not possible without quality radiometric calibration.

Newsletter sign up

Sign up to receive updates directly to your inbox.