[Test] Surveying the condition of a winter wheat culture. What is live NDVI good for?
In our most recent test, we examined the possibilities offered by on-the-go survey and post-process data processing. In this article, we present a possible agricultural monitoring drone survey and its course.
With the latest technical and software developments, the field of application of agricultural drones is becoming wider. The drone and camera manufacturers are making tremendous efforts to provide us with the right technical background. The most important thing, however, is to be able to show tangible results based on the recordings made. It’s not enough to look at beautiful color shots. We need results that we can use. To achieve this, the field must be mapped, an orthomosaic must be made, and this recording must be subjected to professional photogrammetric and statistical analyzes. In the hands of people with adequate expertise, this statistical analysis can provide particularly useful results.
With the solutions of AGRON, the expected yield can be quantified on the basis of the damage caused by wildlife, the germination loss and the plant coverage. The average and per-zone health status of the crop and the nitrogen supply of winter wheat can be assessed, which serve as a basis for application plans.
Many times there is no time to walk around the territory thoroughly, or problems can not be identified from eye level in dense vegetation. The survey can be done immediately next to the field, either by manual or programmed flight.
The protection against pests cannot be differentiated, as they may appear in another after treating one zone. However, the presence of pests and the identification of focal points should take place as soon as possible.
Field zones with major lesions can also be identified with drones equipped with a traditional RGB camera. Then we can purposefully check what we are up against. Let’s show you how!
Immediate surveying options
We recently tested the DJI Phantom 4 Multispectral and, among other things, we examined what the live NDVI, GNDVI, and NDRE features available in the DJI Ground Station Pro, are actually capable of. A programmed flight path was set up for immediate examination of the table (Figure 1). Being a narrow farmland, we determined the flight path in a line at an altitude of 80 meters. At this height, the full width of the area was visible. The cameras were tilted at a 45 ° angle.
Figure 1 Screenshot of the planned flight route in Ground Station Pro
A heterogeneous table on which significant lesions were identified was intentionally selected for the test (Figures 2 and 3). The selected area was surveyed with four settings. Thus, a total of three flights were performed with live NDVI, live GNDVI, live NDRE functions and in comparison, a flight was performed with a conventional RGB camera. The latter was not made with the low-resolution version built into the Phantom 4 Multispectral, but with a 20 MPX X4S camera mounted on a DJI Matrice. The flights were controlled on the same tablet so the drones flew the same route. In the video below, you can see differences in the images of the cameras due to the different viewing angles.
Beyond 10 meters from eye level, the vegetation apparently closes, so the problems are not visible. Although it is a smaller 5-hectare farmland, it would take a long time to walk through it thoroughly. We managed to save significantly on this with a quick aerial visual inspection, which lasted less than 10 minutes for the present field.
Figure 2 There was rooting damage on the spots. The drought affected the entire herd uniformly.
Screen videos of the flights were taken and placed side by side to reveal the differences between the live index and RGB recordings. The most contrasting images were given by live-NDVI and live-GNDVI. However, the coloring of the shots is difficult to interpret. In the case of NDVI, for example, the color scale is completely reversed, so red indicates a healthy plant, while green indicates some problem. The same can be said for GNDVI, where the color scale is also reversed. In the NDRE image, green is healthy and blue indicates the plants with “problems”.
Figure 3 Visible range (left) and live NDVI (right) images from the examined territory
The patchwork of vegetation exposed to severe environmental stress was equally beautifully outlined in the conventional camera shot. The lack of vegetation could only be determined from the RGB recording.
At AGRON, we believe that the color scale carries the primary information that can be achieved through standardization. A properly adjusted color scale clearly shows the lesions. However, the colors and the spots they draw show one side of the coin, the “head,” but the more important side, which shows the value of the coin, remains hidden in the color images. Using multispectral cameras, we can reveal the “tail” side of the coin, i.e. the values assigned to the resulting colors, with which we perform calculations later.
Conclusion: During the surveys, we saw several problems on the ground, so further examinations were needed to quantify the lesions, for which we used programmed mapping.
Analyses with post-processing
We have decision support systems for the post-processing of the recordings from the field. The advantage of post-processing is that the changes experienced next to the table can be accurately assessed, so that further steps can be taken in the light of the results.
Figure 4 Visible range map of the studied farmland. (Vizuális felmérés – Visual monitoring)
In the case of the damage caused by wildlife, the extent of the harm can be determined with an accuracy of 0.5%. We can determine how much yield loss can be caused by rooting damage and based on this we can decide whether it may be worthwhile to claim compensation. In the case of green vegetation, the current expected yield can be summed up based on the utilization of the cropland and the previous yields of the given crop. After the exclusion of plant pathological changes, the nitrogen supply of winter wheat can be measured. After the identification of nutrient-deficient spots, the given zones can be treated differently, for example with foliar fertilizer. By determining the health status of the board, can it be decided whether or not the use of a plant conditioner is necessary?
The AGRON Maps system, in which orthomosaic concatenation is basically free, provides a professional basis for making such decisions. The tests that can be performed on the completed orthomosaics are based on the price per hectare, so the costs associated with the survey can be planned in advance per territory. Based on the preliminary surveys on the platform, we conducted a plant coverage survey and a stress status survey.
The plant coverage survey does not provide information on the physiological state of the vegetation, it only shows the “greenness” of a given area (Figure 5). With this analysis, the cultivating traces cannot be distinguished from the actual damage, so in fact the average coverage or utilization of the field can be assessed with it. However, the result provides good information of the expected yield in the light of the average yield of the given crop measured in previous years. In the case of plant coverage survey, an error of 2-3% is to be expected due to the traces of cultivation, so its use as a damage assessment is for information purposes only. In the present field, the coverage proved to be 93%, so based on this, the previously mentioned rooting damage can be around 5%. However, the study is not a substitute for a much more accurate wildlife damage survey developed specifically for this purpose.
Figure 5 Utilization of the examined farmland (Növényfedettség felmérés – Plant Coverage survey; A tábla átlagos fedettsége – Average coverage of the field; Fedettség – Coverage; Talaj – Soil; Alacsony – Low; Közepes – Medium; Magas – High)
Stress status surveys can be used to demarcate healthy and environmentally exposed zones. The health status of the vegetation is classified on a scale of 1-10, from which an application map can also be designed. The average health status of this territory was rated 4, which is a pretty low value. The cause is the drought damage mentioned earlier and the rooting damage. The crops would benefit from some rainfall, but it is recommended to apply foliar fertilization before flowering. The board can basically be divided into two large zones, so that the treatment can also take place in two doses.
Figure 6 Stress assessment of winter wheat crop sown on the field (Stressztérkép – Stress map; zónák – zones; Egészségügyi állapot – Health status; A tábla átlagos egészségügyi állapota – Average health status of the field)
We could see how the live image recording solutions help with rapid diagnostics of the cropland. Focal points and problem zones can be found quickly on the territory. Fortunately, diagnostics can also be performed with a much cheaper drone equipped with an RGB camera, so there is no need to invest in much more expensive technology for a live image index.
However, in order to quantify the changes in the culture, studies should be performed by post-processing in order to have the most accurate information available about our farmland.
For basic physical surveys, drones equipped with an RGB camera are sufficient, but are highly recommended for wildlife damage, weed and plant number surveys (due to the need of high resolution for these surveys). However, the physiological examination of the vegetation and the application plans that can be prepared on the basis of the survey must be based on preliminary data collection.
AGRON Analytics Ltd.
All rights reserved!