The use of drone technology has brought many changes to society as a whole, transforming our lives and the way we do business. The agribusiness sector embraced drones in a way that transformed modern agriculture through monitoring, pesticide application and, more importantly, image capture.
Drone applications in agriculture
Spraying and fertilizing
Drones have a potential use in pesticide spraying, however, there are still legal barriers to this practice in Brazil. The biggest barriers are related to environmental problems and anti-terrorism laws, which prevent drones from carrying chemicals. However, the use of drones in pesticide application brings several benefits, such as:
– It does not cause soil compaction (increased soil density, which hinders, for example, the growth of roots. It is usually the influence of tractors and other land farm machinery);
– There is no need to have a person (applicator) exposed to the pesticides during application. Thus, possible contamination by the pesticides is avoided;
– It allows taller plants to receive pesticide application;
– It reaches hard-to-reach terrains;
– It reduces costs;
– It reduces environmental risks.
Precision fertilization planning
Nitrogen-deficient areas can be easily identified through the use of drones with cameras and sensors. Sensors are calibrated to limit the effect of the sunlight and, thus, allow an accurate calculation of the green area.
Throughout the production cycle, hundreds of images of crop development are extracted. All images are placed side by side to generate a map. Based on these images, software can be used to identify growth patterns in the crops. Thus, a fertilization program can be designed to adjust the nutritional requirements of the crops.
Disease and weed control program
Drones can help identify specific properties — such as foliage coloring and water balance, among others — of plant species and crop areas that have succumbed to diseases. The images captured by the drone, through powerful software, enable this data to be analyzed and can direct measures to control the crop, both in relation to weed management, as well as disease control.
Mapping and differentiation between trees and terrain
Fruit growers and foresters can benefit from plant records and spacing with calculations of the land to be cultivated. Drones enable access to remote locations that would be difficult to cover. Hundreds of hectares can be mapped daily, identifying problems on the ground with accuracy of up to 10 cm. With this data, it is possible to create a 3D model in order to have a better analysis of crops.
Relationship between the use of drones and images in agriculture
As can be seen in the examples, most of the tasks in which drones can make a difference in agribusinesses are related to the capture and treatment of images of crops. Image analysis is not something new among the technologies used in precision agriculture, especially the use of images from satellites, which have been used in agriculture for a long time.
The use of this type of image, however, has some restrictions, such as: high price of images; data analysis only in a macro way, since glebe details may not have very good granularity; delay in obtaining images; and the need for payment, as some type of subscription is usually required to have access to such services.
The use of drones in precision agriculture has grown a lot in recent years, as an option that offers answers to satellite imagery restrictions. UAVs (Unmanned Aerial Vehicles, another name for drones) are a revolutionary tool for farmers to gain knowledge about crop development and to grow productivity and maximize production. Drones have the unique advantage of providing quick responses with field data, as well as providing images with a very high level of detail. Other than that, with them, it is possible to get images of many hectares with a simple flight and with more affordable costs to rural producers.
Drones capture images, but they need powerful software to perform their analysis. A good drone system in agriculture needs to encompass drones, sensors and software. The use of software in drones allows, first of all, fields to be mapped and flight plans to be created. After the drone follows the flight plan correctly, it is necessary to apply another software, which will define the actions to be taken and send data to the team and servers.
The sensors are an important part of this set, as they capture extra data such as temperature, solar radiation and air humidity, among others, helping to enrich the information provided to the system. The use of the images obtained by the drones would allow the collection of data from crop stems, with analysis of vegetative indexes (algorithms used to analyze images of crops based on color saturation) of the batch, analysis of pest and disease data.
Therefore, the use of drones in the fields enables image analysis with specific details of the agricultural property, with more detailed data. Thus, they allow the gathering and processing of agricultural and livestock production data in a more focused way — unlike satellite images, which are much more generic — and enable the comparison of more detailed data, such as obtaining samples of pests infestations or even diseases in a field.
Image analysis in Agribusiness, coupled with the latest software technologies — such as machine learning, artificial intelligence, localization, event detection, and other technologies — enables precision agriculture to take a leap in information quality and details, which brings improvements to decision making and allows increased agricultural productivity.
An example would be a coffee plantation. The geo-located images of a coffee plant captured by a drone already bring, by themselves, a great possibility of assistance in crop management, through the analytical eye of the producer, who reads the images. But, in addition to the human eye, machine learning algorithms can also be implemented in image analysis. These algorithms can be used to signal to the producer the occurrence of unexpected events, such as water deficiency, nutritional deficiency or, even, sick plants.
In perennial cultures such as coffee, image analysis can also store data from historical series and, therefore, allow early analyses based on the coffee plantation’s own history. Image analysis would allow the state of vegetation before fertilization and its evolution to be analyzed.
In addition, collected data can be used to refine software analysis. For example, during the winter, the shade of green of foliage is deeper and, during the summer, less intense, without this being a problem. However, most software would be calibrated to identify this variation as an indication of disease.
With image data and historical series data production, it would be possible to make production analyses based on crop characteristics and verify changes consistent with the period of the year (e.g. the effect of fertilization on the plantation). Thus, the uses of image data and their analyses will depend on the information and needs of the farmer/producer.
Vegetation indexes for image analysis
There are several vegetative indexes. These indexes are algorithms developed to analyze the images captured by drone cameras and to offer various reports comparing data through the coloring of each index. The indexes are identified through image analysis and enable the verification of plant health data in precision agriculture.
Regardless of the purpose, all vegetative indexes should reflect the reality of the plants in the field. The results of the vegetative indexes of plants can vary greatly from culture to culture. Thus, the recommendation is to use the vegetative index that best represents the agricultural culture that is the focus of the analysis and that provides the most relevant information — such as the vegetation status, water balance, the need for nutrition, among others.
NDVI (Normalized Difference Vegetation Index)
NDVI (Normalized Difference Vegetation Index) is currently the most used index in agriculture, due to its versatility and reliability in the analysis of biomass data in general. NDVI quantifies vegetation by measuring the difference between the red and infrared levels in the images. The difference between the values of green and red coloration of the plants differentiate the plants from the soil. It is used to detect differences in a crop’s coverage — which areas are cultivated and what is the plant density there —, emphasizing the green of a healthy plant.
However, the red light used by NDVI is generally absorbed at the top of the coverage. Thus, lower levels of vegetation do not appear as strongly in images processed by NDVI. The greater the amount of leaves (trees, corn in its final stage, for example), the greater the possible distortions.
Grass and cereals in their final stages of production pose another problem. At these stages, these plants become saturated with chlorophyll. This makes it difficult to analyze color variation, as it may appear that the vegetation is in trouble (high red levels), but this variation is part of the natural cycle of grass and cereals.
ENDVI (Enhanced Normalized Difference Vegetation Index)
ENDVI is a close equivalent of NDVI, but it uses blue and green light instead of just red. The ENDVI algorithm is better at isolating plant health indicators, since the absorption of blue light and high reflectance of the color green are reliable markers of plant health. Thus, in cases where the main focus of the images analysis is the health of the plants, better than the vegetative index, the ENDVI method, as it has more resources to differentiate states of each plant.
VARI (Visual Atmospheric Resistance Index)
VARI is a vegetation index that was originally designed to be used for satellite images. Roughly speaking, VARI indicates “how green plants are”. This vegetative index is based on the presence of blue in the spectrum calculation.
When sunlight reaches Earth’s atmosphere, light is spread in all directions by gases and particles in the air. Blue light tends to spread more than all other colors, as it is emitted at wavelengths shorter than the rest of the visible color spectrum.
VARI does not dispense with the use of other indexes, such as NDVI, but it is a great aid in cases where the atmosphere does not have a great impact on the images — as is the case with images captured by drones, since they are taken much closer to the field and the atmospheric effect is very small.
NDRE (Normalized Difference Red Edge)
NDRE is an index used to evaluate the chlorophyll content in plants, as well as their nitrogen uptake and the demand for fertilizers. It uses the band at the edge of the red end of the light spectrum, so high NDRE values represent a greater amount of chlorophyll. Soil images have lower values, weakened plants, intermediate values, and healthy plants have higher values.
Thus, the use of NDRE allows for a more precise analysis of the transition between the healthier plants to the less healthy ones, as it does not saturate the map as easily as NDVI. Larger plants, of greater density, can be analyzed through NDRE, since this index allows a more detailed analysis of the plant as a whole, unlike NDVI, which focuses on the top of plants. With smaller, less dense plants, NDVI can be used.
Vegetation indexes have been used for more than 40 years by scientists and agronomists to assess the health of crops and plants. That is, it has been a long time since crop images have been analyzed by indexes such as NDVI. However, with the arrival and adoption of drones in Agribusiness, what is noticeable is that image analysis technologies can jump to a new level of usage in a crop’s productivity. Combining technology of image analysis, drones, sensors and software, precision agriculture has a much more effective role in helping agribusinesses.
Satellite image processing has already shown its usefulness in agriculture. However, factors such as cost, representation of only large areas, small granularity (compared to the degree of detail that drone imagery allows) and the fact that satellites capture only static images show us that the use of drones and the images they are capable capturing provide a new level of information and guidance for precision agriculture.
Drones can generate clearer images, with geographical location and with a high degree of reliability, in addition to allowing routes to be traced and planned in order to obtain data. This data, along technologies such as machine learning and artificial intelligence, allows increasingly more accurate and relevant data for decision-making by rural producers.