HOW TO PERFORM VEGETATION ANALYSIS WITH A SINGLE CAMERA

LUMENERA WHITE PAPERSOLUTION SERIES SHEET

HOW TO PERFORM VEGETATION

ANALYSIS WITH A SINGLE CAMERA

When light interacts with any object, certain wavelengths are absorbed, reflected, or refracted. This is
especially true when it comes to plants because they draw a significant portion of their energy through
photosynthesis by absorbing the sun’s energy and converting it into glucose. Through years of studying
vegetation, a number of methods have been developed to monitor plant health using their reflected spectral
signature.

These methods have been broken down into over 150 vegetation indices (VIs), some of which include the
Simple Ratio Index (SR), the Atmospherically Resistant Vegetation Index (ARVI), the Vogelmann Red Edge
Indices (VOG1, VOG2, VOG3), and the Photochemical Reflectance Index (PRI). The Normalized Difference
Vegetation Index (NDVI) is arguably the most common and well known vegetation index.

LUMENERA SOLUTION SHEET

www.lumenera.com 1
 HOW TO PERFORM VEGETATION ANALYSIS WITH A SINGLE CAMERA
LUMENERA WHITE PAPERSOLUTION SERIES SHEET

NDVI - EXPLAINED
NDVI is based on the known reflectance from both
healthy and unhealthy vegetation. As described in Results from the equation will span from -1 to 1, where
the graphic below from the February 2012 edition inorganic material will typically have negative values
of Applied Optics, the red data points show the and healthy vegetation will be between 0.2 and 0.8.
reflectance of unhealthy vegetation and the green
data points show healthy vegetation.

Figure 1. Reflectance of Healthy and Stressed

Vegetation in Relation to Wavelength
Source:1
The spike at 550 nm in the healthy response curve
translates to the green color associated with healthy Figure 2. Example of NDVI Calculation for Healthy
plants. In other words, they reflect more green light and Stressed Vegetation
Source: Wikipedia
than blue or red because these wavelengths are
absorbed by the plant during photosynthesis. Healthy Since computers typically display images with 8-bits
plants also reflect a high amount of infrared light to of color depth, a lookup table (LUT) is applied to the
protect themselves from overheating. results of NDVI calculations to visualize the data.
Therefore, it changes the scale of the images from
The principle of NDVI is to compare the ratio between between -1 and 1 to between 0 and 255. There are a
large number of LUTs that exist for NDVI images, so it is
the red (~650 nm) and near infrared (~850 nm) light
important to ensure that the same one is applied when
reflected by the vegetation. The ideal method to
comparing images. Below is an image courtesy of
accomplish this is through the use of two monochrome Kansas State University showing an RGB and
cameras with filters allowing each of them to see false-color NDVI image of a soybean plant where blue
only the specified wavelength. The images are then indicates a low NDVI value and red indicates larger
correlated and the following NDVI formula is applied: NDVI values.

2
1
Justin A. Hogan, Joseph A. Shaw, Rick L. Lawrence, and Randal M.
LUMENERA SOLUTION SHEET
Larimer, “Low-cost multispectral vegetation imaging system for
www.lumenera.com detecting leaking CO2 gas,” Appl. Opt. 51, A59-A66 (2012)”
 HOW TO PERFORM VEGETATION ANALYSISDesigning
WITH A SINGLE
a Vision
CAMERA
System
LUMENERA WHITE PAPER SERIES SHEET
SOLUTION
White Paper

DEFINING SENSITIVITY

Figure 4. Typical Quantum Efficiency Curve

Figure 3. Soybean Plant in Natural and False-Color for a Color Sensor
NDVI Representation
Source: K-State Flickr It is important to note that this is not the only solution

TIP
APPROACH TO SINGLE-
for creating a single-camera NDVI system and that
other filter setups exist.

CAMERA NDVI
With advancements in UAV technology, NDVI is
becoming more accessible to larger groups of people
at lower costs. A major issue concerning UAVs is their
payload requirements where size and weight need to
be minimized as much as possible to prolong flight
time. Knowing the typical spectral response of a color
image sensor, seen in Figure 4, it is possible to use
filters to target specific color channels and turn a
single camera into a specialized NDVI camera yielding
a close approximation to a two-camera system.

Typically, color cameras have an NIR-cut filter to Figure 5. Spectral Response of the Dual Bandpass
negate the camera’s responsivity past 700 nm. Filter Used in This Solution Sheet
Removing this filter and adding a dual bandpass filter
Since this filter specifically targets blue and NIR
to observe the vegetation’s reflectance at two specific
wavelengths, modifications need to be applied to
wavelengths allows for an approximation of NDVI. For
the standard NDVI formula. In the case of single-
the purpose of this solution sheet, a dual bandpass
camera NDVI with this particular filter, the camera’s blue
filter with wavelengths of 475 and 850 nm, or blue and
channel measures the vegetation’s absorption of light
NIR, was used and its response can be seen in the
and its red channel is used to measure the reflected
Figue 5.
NIR light.

LUMENERA SOLUTION SHEET

www.lumenera.com 3
 HOW TO PERFORM VEGETATION ANALYSISDesigning
WITH A SINGLE
a Vision
CAMERA
System
LUMENERA WHITE PAPER SERIES SHEET
SOLUTIONWhite Paper

Keep in mind that the blue channel is also

measuring some NIR light, so a correction for this is
required. As the following graph shows, the NIR
response of a typical image sensor is very similar in
both the blue and red channels.

Depth of Field:
This equation could be further adjusted based on
Range of distance in front of the camera the sensor response by adding a gain factor to red.
where the target remains in focus. This could be calibrated using a two-camera solution
and experimental results. A more general equation
including the determined gain factor (k) would be:

Figure 6. Sensor Response With Dual Band Filter

Attached
Therefore, the data obtained in the red channel can be
subtracted from the blue channel to compensate for
CREATING A SINGLE-
the blue channel’s sensitivity to NIR light. Knowing this, CAMERA NDVI SYSTEM
an approximation to the NDVI equation can be derived
as seen below: Creating a single-camera NDVI system is
relatively simple and possible with readily available
R’ is the information registered by the camera’s red “off the shelf” components. All Lumenera cameras are
channel; available without the IR cut filter which means that
the entire portfolio of cameras is available for this
And, B’ is the information registered by the camera’s solution. The camera will need to be paired with a
blue channel. dual bandpass filter allowing it to monitor at least one
absorption band (blue or red) and the high reflectance
Using the red channel (R’) to measure the reflected band (NIR) of the vegetation. For this document, the
NIR light and the blue channel (B’) to measure the Lumenera Lt965RC-WOIR was paired with MidOpt’s
vegetation’s absorption, the NDVI approximation dual bandpass DB475/850.
equation becomes:
An image of the following scene containing (left to
right, back to front) two plants, an orange, a glass of
water, the filter calibration target, an apple, a water
bottle, a banana, and some cashews was taken using
the camera and filter mentioned above.

LUMENERA SOLUTION SHEET

www.lumenera.com 4
 HOW TO PERFORM VEGETATION ANALYSISDesigning
WITH A SINGLE
a Vision
CAMERA
System
LUMENERA WHITE PAPER SERIES SHEET
SOLUTIONWhite Paper

The captured image, seen below, was then imported to The image was then processed using an open
FIJI (ImageJ) – an open source photo editing software. source plugin for FIJI available on Public Lab called
Photo Monitoring Plugin, written by Ned Horning.

The first processing step was to calculate the

image calibration coefficients based on the calibration
target and data provided by MidOpt. An ROI was
drawn over each of the three reflectance targets and
their associated reflectance data at both bandpass
wavelengths was input to the plugin. At this step, the
red channel was subtracted from the blue channel to
account for the blue channel’s sensitivity to NIR light.

The second processing step then calculated the

image’s approximate NDVI using the calibration
coefficients. The camera’s blue channel was selected
for the visible band and the red channel for the NIR
band. The below false-color NDVI approximation
image was then generated from the plugin.

Figure 7. RGB View of Example Scene

Figure 9. False-Color NDVI Approximation Generated

by the Photo Monitoring Plugin Using the
NDVIBlu2Red LUT
Through the approximation, all organic matter is
BIT DEPTH highlighted by a yellowish-green color. It is important
to note that the reason the underside of some of the
Figure 8. View of Example Scene Through DB475/850 leaves is blue (indicating a negative NDVI) is because
Filter Using Lt965RC-WOIR only the upper side of the vegetation is meant to reflect
infrared light.

LUMENERA SOLUTION SHEET

www.lumenera.com 5
 HOW TO PERFORM VEGETATION ANALYSISDesigning
WITH A SINGLE
a VisionCAMERA
System
LUMENERA WHITE PAPER SERIES SHEET
SOLUTION
White Paper

Furthermore, the inorganic bottle cap has positive Core competencies include digital bus technologies
NDVI values due to the high level of infrared reflectance such as USB 3.0, USB 2.0, Ethernet, HDMI, and
of the cap compared to its absorption of blue light. Gigabit Ethernet (GigE) as well as a complete
command of digital imaging hardware and software
CONCLUSION built around CMOS and CCD based imagers. Our
diversity provides our customers with the benefits of
NDVI is arguably the most popular vegetation superior price-to-performance ratios and faster time-
index and is being more widely adopted due to to-market.
advancements in UAV technology. Since UAVs have
limited payload capacities, single-camera NDVI
solutions are being explored and adopted as suitable
approximations of the traditional dual-camera systems.
Removing a color camera’s IR cut filter and pairing it
with a dual bandpass filter allows the system to strictly
observe the absorption and reflectance bands
associated with NDVI. Lumenera’s entire portfolio
of cameras is suitable to be designed into a single-
camera NDVI system and can easily incorporate third
party filters into the camera’s design, such as those
offered by Midwest Optical Systems and highlighted in
this document.

If you have questions about camera specifications for Lumenera Corporation

your aerial imaging payload, reach out to our imaging 7 Capella Court
experts at [email protected]. Ottawa, ON
Canada, K2E 8A7
ABOUT LUMENERA 613-736-4077
[email protected]
Lumenera Corporation, a division of Roper
www.lumenera.com
Technologies, headquartered in Ottawa, Canada,
is a leading developer and manufacturer of high
performance digital cameras and custom imaging
solutions. Lumenera cameras are used worldwide in
a diverse range of industrial, scientific and security
applications.

As a global market leader Lumenera provides an

extensive range of high quality digital cameras
with unique combinations of speed, resolution, and
sensitivity to satisfy the demands of today’s imaging
applications. Lumenera also offers custom design
services to OEM partners requiring specialized
hardware and software features.

LUMENERA SOLUTION SHEET

www.lumenera.com 6