IEEE Geoscience and Remote Sensing Magazine - June 2019 - 186

conversion of the cameras later in this section). The first
task was to reduce the complexity of satellite images by
applying one of many indices. Because our program deals
with vegetation, the index of choice was the Normalized
Difference Vegetation Index (NDVI) [15], which is one of
the most common and robust indices. It is calculated according to
t NIR - t RED
NDVI = t NIR + t RED ,

(1)

where t NIR is the reflectance in the near-infrared spectral
range from approximately 800 to 900 nm, and t RED is the
reflectance in the visible red range from roughly 630 to 690
nm, depending on the satellite's sensor system. NDVI values
range from +1 to -1. Healthy plants have high NDVI values
of 0.4 to approximately 1.0,
soil has values close to 0, and
water typically has negative
OUR PROGRAM WAS
NDVI values [9].
EDUCATIONAL AND SUBJECT
Recent years have given rise
TO BUDGET CONSTRAINTS,
to the idea that NDVI meaSO WE SOUGHT A DO-ITsurements can be acquired using low-cost infrared-convertYOURSELF SOLUTION TO
ed digital cameras mounted
PROVIDE STUDENTS WITH
on tethered balloons, kites,
CAMERAS SUITABLE
or unmanned aerial vehicles
FOR ACQUIRING REMOTE
(UAVs) for application by farmSENSING DATA.
ers, for scientific studies (e.g.,
[16]), or for educational purposes at the high school, college, and university levels [18]. Despite the promise of the
idea, researchers [11] found that standard red-green-blue
(RGB) cameras outperformed the modified NDVI cameras
both in a controlled laboratory setting and in a less-well-controlled field setting with short distances between the cameras and the investigated plants. In contrast, other researchers
[4], [13], [14] found some advantages to using the technology
for NDVI measurements from balloons or UAVs. The conflicting results were summed up by researchers [4]: "At this
stage, this study can only suggest that the sensor might be
used for support of visual paddock assessment and broad-

(a)

(b)

band vegetation indices." Nonetheless, some companies are
selling UAV-mounted NDVI cameras for precision farming
(e.g., https://sentera.com; https://www.mapir.camera; http://
www.agrocam.eu).
Our program was educational and subject to budget
constraints, so we sought a do-it-yourself solution to provide students with cameras suitable for acquiring remote
sensing data. A side benefit was that this enabled students
to learn about camera techniques by hands-on experimentation. We therefore relied on an online tutorial (https://
publiclab.org/wiki/infragram-point-shoot) that shows how
to convert a simple point-and-shoot camera into an NDVI
camera. Basically, we replaced the infrared-cutoff filter in
consumer-grade digital cameras (which have sensors that
are sensitive to infrared radiation) [10] with a cheap, lowpass, red-gel filter that facilitates band separation according
to the NDVI (Figure 1).
The low-pass filter overlaps the built-in Bayer RGB filter of the camera sensor. Thus, only radiation with wavelengths higher than approximately 580 nm can reach the
sensor [17]. In images acquired with the converted camera,
the blue channel represents near-infrared radiation from
approximately 800 to 900 nm, and the red channel represents visible red and some infrared radiation. On images
taken with the simple NDVI camera, vegetation is represented by bright blue colors: the strong infrared reflection
of healthy plants is detected on the infrablue image channel (Figure 2).
The images can be analyzed according to the NDVI with
any kind of geographic information systems (GIS) software,
using the blue image channel for t NIR and the red channel
for t RED . The derived NDVI is subject to a systematic bias,
resulting in NDVI values that are too low compared with
real satellite data. One reason for the low NDVI values is
that the red image channel also detects some near-infrared
radiation, so the reflectance from the red channel is overestimated. The validation for the experimental protocol was
provided by the strong correlation (R 2 = 0.97) between
data from our low-cost NDVI camera and the NDVI values derived from measurements with a professional field
spectrometer (ASD FieldSpec 3) in a field measurement

(c)

(d)

FIGURE 1. The conversion of a Mobius ActionCam camera. (a) The camera is opened. (b) The built-in infrared-cutoff filter is removed and
(c) a circular piece of red gel filter is cut out. (d) The completed conversion. The only tool required for executing the modification is a small
screwdriver.

186

IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

JUNE 2019
https://www.sentera.com https://www.mapir.camera http://www.agrocam.eu http://www.agrocam.eu http://publiclab.org/wiki/infragram-point-shoot http://publiclab.org/wiki/infragram-point-shoot
IEEE Geoscience and Remote Sensing Magazine - June 2019

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - June 2019
