How to Choose a Survey 3 Camera Model

We provide 6 different filters for the Survey3 camera: OCN, RGN, NGB, RE, NIR and RGB. The filter captures 3 channels of light information as follows:

OCN filter (orange + cyan + NIR):

Red Image Channel = Orange Light

Green image channel = cyan (blue/green) light

Blue Image Channel = NIR (Near Infrared) Light

The OCN filter is an improvement over our RGN as it increases contrast within vegetation and reduces soil noise. It is better to use OCN if your vegetation has soil in it, and RGN if the crop has a more solid canopy (less soil pixels). It can be used with NDVI index like RGN filter. Read here to learn more about its advantages over RGN filters. If you are looking for the best camera to buy for measuring general plant health, then the OCN models are the ones for you.

RGN Filter (Red + Green + NIR):

red image channel = red light

Green image channel = green light

Blue Image Channel = NIR (Near Infrared) Light

The RGN filter used to be our most commonly purchased model, mainly due to its ability to capture the red and NIR wavelengths required by the popular NDVI index (see below for more info). NDVI is often used as a general plant health and vigor index, basically it shows which areas are the healthiest compared to those which are not so healthy. Our new OCN filter generally gives better results, you can read more details here.

NGB filter (NIR + green + blue):

Red Image Channel = NIR (Near Infrared) Light

Green image channel = green light

Blue Image Channel = Blu-ray

The NGB filter is often used in the ENDVI index, which is basically an enhanced NDVI. It takes into account a plant's green reflectance when determining plant health, rather than just using reflected near-infrared (NIR) light as used by the NDVI index. Some applications (such as DroneDeploy) do not allow you to calculate ENDVI, so be sure to check which indexes are supported. You can also calculate the NDVI index using blue vs. NIR light, which may show different results compared to using an RGN camera. The best way to think about the difference between RGN and NGB models is that most of the time the RGN model is the better choice, but the NGB model may show you things that the RGN can't, so if your budget allows, use both camera and recommends comparing the results.

RedEdge Filter (RE):

Red image channel = RedEdge (725nm) light

Green image channel = not used

Blue image channel = not used

The RedEdge filter is used to capture a single band of reflected light in an area called the red edge. This region around 700-800nm is where plants have different reflectivity, which is closely related to their health. Plants that reflect more red-edge light are generally healthier than those that don't. When processed with our MCC application, the output image will be a single image band, i.e. black and white. White pixels have a high red edge reflectivity, and black pixels have a low red edge reflectivity. You can ignore the green and blue image channels because they don't contain useful data compared to the red channel.

Near Infrared Filter (NIR):

Red image channel = near infrared (850nm) light

Green image channel = not used

Blue image channel = not used

Near-infrared filters are used to capture single-band reflected near-infrared light. When processed with our MCC application, the output image will be a single image band, i.e. black and white. White pixels will have high NIR reflectivity, while black pixels will have low NIR reflectivity. You can ignore the green and blue image channels because they don't contain useful data compared to the red channel.

RGB filter (Red + Green + Blue):

red image channel = red light

Green image channel = green light

Blue Image Channel = Blu-ray

RGB filters are the typical filters installed on all cameras, and they capture colored light the way our eyes see the world. RGB cameras are often used in conjunction with multispectral cameras to provide a reference image to the viewer. This referencing is often necessary to relate what our eyes see to what a camera capable of capturing near-infrared (NIR) light sees.

Multispectral Index Formula

Once the OCN, RGN, NGB, RE and NIR model cameras have captured images, they should be calibrated using our calibration targets. Once calibrated, the images can be stitched in the program of your choice, such as Pix4D, Agisoft, Drone Deploy, Agribotix, MapsMadeEasy, Simactive, Icaros, and more.

Many of these applications provide so-called raster/index calculators, which perform mathematical operations on the pixels of an image. The pixel values resulting from computing an index represent the range of pixels that depend on the index and what it computes. Many programs make this calculation a one-click process, but let's explain it in more detail:

Let's calculate the popular NDVI index using the RGN filter as an example:

As you can see in the formula above, the NDVI index uses NIR and red light. So, for an RGN filter camera model, this would be the blue image channel (NIR) and the red image channel (RED). The handler will take the pixel values in the red and blue image channels and plug them into the above equation. The resulting pixels will all have a value from -1 to +1. For plants, real plants have NDVI values between about 0.2 and 0.8. We then apply a color lut to the pixels so our eyes can interpret the data more easily. The color luts are the ones you've probably seen before from green to yellow to red (high health to low health).

Important note about reflection calibration using our ground targets:

The pixel values mentioned above tell you whether you are looking at healthy vegetation or if nearby dirt is affected by the calibration procedure. If you don't calibrate the image, the resulting values are usually negative and basically garbage. The resulting color lut picture, often called a "pretty picture", may show a similar green-to-yellow-to-red map, but without calibration you won't be able to compare one field location to another, one ambient lighting condition Compare to the results of another ambient lighting condition, or basically from moment to moment (week to week, month to month, etc.). Without calibration, you're not aligning pixel data to a known standard, so the values won't be comparable.