What can RGB and multispectral cameras be used for?
In previous articles We have already discussed the plant physiological aspects and practical applications of drone technology. Now we collect which cameras and which analyzes can be applied for professional scouting of the vegetation.
The basis of the analyses: orthomosaics and reflectance maps
The basis of all drone data collection is to cover the area to be examined with photographs during autonomous flight. These photo packages are the basis for future studies.
Therefore, it matters if the area is photographed with a conventional (RGB) camera or a multispectral camera. However, we can’t say that one is better than the other because we can use both technologies, just not for exactly the same purpose.
During photography, in both cases, we need to cover the entire area and leave specific overlaps between them so that the system can concatenate them during processing. The concatenated images are called orthomosaics, a version of which is the reflectance map created by multispectral cameras. While a map made from a photo package taken with a traditional camera can be interpreted at a glance, photos taken by multispectral cameras measure the reflectance of a range of light from the vegetation, which can be interpreted in further steps.
Traditional cameras provide information on wildlife damage, elemental damage, sowing failure, growth problems, weeds and plant counting. Multispectral cameras show the intensity of the reflections measured at different wavelengths, the physiological state of the vegetation. Both methods provide important and usable results for the field work.
Figure 1. Trampling damage in flowering (left) and mature (right) rapeseed
Figure 2. Young sunflower suitable for plant counting (left) and sowing error in maize (right)
We can examine the red, green, blue, near-infrared, far-red ranges, which are intended to serve as the basis for later indexes (e.g., NDVI, NDRE). Thus, the reflectance map, which can be created using multispectral cameras, contains the values of the given wavelength ranges per pixel. Ennek megfelelően képpontonként képesek vagyunk matematikai műveleteket végezni. Az egyenlet határozza meg a készülő térképünk minden egyes képpontját. However, after performing the operation, we still do not get to the well-known colored green-yellow-red area, but to its base, a grayscale index map. This map contains important values, as each pixel from black to white is a normalized value.
In general, black pixels indicate areas where there is no vegetation, gray areas indicate weak vegetation, and pixels approaching white indicate dense, photosynthesizing vegetation. Of course there are many, many 100 or 1000 steps between the white and black color ranges, depending on how many bit photos we took.
It is worth standardizing the coloring methods, so as the values of an index map change, the color scale also changes in a well-defined way. Time series maps can be easily created to provide accurate information throughout the vegetative period of a culture. In this way, the colors green-yellow-red also make sense. The last step is to make it much easier for the human eye to understand what we are actually seeing, so we match the pixels as follows (simplified):
- black parts – shades of red
- grayish parts – shades of yellow
- white parts – shades of green
This is how our red-yellow-green NDVI map was born. Of course, the NDVI formula was used for illustrative purposes only, as it is the best-known index among the hundreds that exist. Most indexes combine the near-infrared range with some other color range, but we even utilize the far red band, which we can use to calculate additional indeces. These maps will be the basis for further analysis steps!
Figure 2. Grayscale index map (left) and NDVI map (right) of a winter wheat field
There is also a solution that can seemingly create an NDVI map based on basic RGB photos. However, this is only a “mock NDVI” as we cannot use near infrared band for it. In this case, we can measure the “greenness” of the area, so the vegetation can be separated from the soil. Thus, a non-real NDVI map is obtained, but in many cases this may also be sufficient to assess the utilization of the table. However, it is not suitable for accurate vegetation analyzes.
Vegetation analysis workflow
- RGB cameras can be used to measure the physical condition of vegetation.
- Accurate physiological survey and zoning can be performed during the examination of cultures.
- The system automatically creates your orthomosaics.
- The different channels (wavelengths) of light reflected from the leaves of the plants are examined.
- Each channel bounces off plants and other landmarks in different ways.
- We perform mathematical operations on a pixel-by-pixel basis, from which the index map is born.
- The index map is colored to give you a better insight.
- Statistical analyzes are performed on the completed index images to perform the desired studies.
- We summarize the results in the form of a report and an application map.
The number of color channels in the base map is determined by the camera in each case. As mentioned above, a basic camera sees R / G / B colors. Agricultural cameras are capable of examining additional channels. These include the near-infrared (NIR) and far-red (RedEdge) bands. These two wavelength ranges are the most sensitive to photosynthesizing plants. They are typically called multispectral cameras. However, there are also cameras that can detect the near-infrared range after filter replacement. However, in this case, we lose some of our base band from the RGB. Typically, near-infrared appears in the “spot” of blue after the device is capable of receiving this wavelength range.
DJI Phantom 3, Phantom 4, Phantom 4 pro/advanced, Mavic Pro, Mavic 2
Usually, the drones available for purchase have their own RGB color composition. In this case, we can create traditional – visible range maps, on which e.g. wildlife caused damage, storm damage, water pressure, plant height and green area survey is possible.
Camereas with modified lenses
DJI Phantom 4 agro (ndvi) mezőgazdasági drón, MAPIR Survey sorozat, Sentera Single NDVI stb.
The near-infrared channel is displayed, making it suitable for measuring true NDVI. It maps photosynthetic vegetation with high efficiency. These cameras typically have 3 channels. The near-infrared (NIR) channel arrives in place of either the red or blue channel. They do not have an incident light sensor, so consecutive tests are not compatible. In return, however, the cameras are cheaper and still suitable for basic diagnostics.
Parrot Sequoia/Sequoia+, Micasense RedEdge M/MX/Altum, Sentera 6X, DJI Phantom 4 Multispectral
These cameras specifically designed for professional agricultural use. They have a separate sensor for each channel, so when taking a photo, as many images are captured as the range of the camera can detect. This allows you to map the area by channel. It is a suitable tool for any test that may be needed in agriculture. Thanks to the incident light measuring sensor, the calibration is carried out according to the light conditions, the results are absolute values, so they can be compared with later results. They are suitable for mapping early problem focal points and determining nitrogen supply using the distant red band.
Analyes for each camera types
Analyses available in AGRON Maps:
- Traditional RGB camera: Wildlife damage , Plant coverage, Terrain map, Plant development status, Weed survey, Plant counting, Visual monitoring
- Cameras with modified lenses: Vegetation diagnostics, Plant coverage, Terrain map, Plant development status
- Multispectral cameras: Plant stress surves, Plant coverage, Terrain map, Plant development status, Visual monitoring (In case there are R,G,B bands), The resolution of these cameras are significantly lower than that of RGB cameras as plant physiological surveys do not require an accuracy of 2 cm. Although higher resolution tests can be performed at low flight altitudes, they are not recommended due to the drastically increased flight time.
Application plans based on camera types
The application plans are based on the analyses that can be ordered in AGRON Maps:
- Traditional RGB camera: Differentiated sowing plan, Differentiated weed control plan
- Cameras with modified lenses: Differentiated sowing plan, Differentiated weed control plan, Differentiated nutrient application plan
- Multispectral cameras: Differentiated sowing plan, Differentiated weed control plan, Differentiated nutrient application plan
AGRON Analytics Ltd.
All rights reserved!