Potential of UAV-derived RGB spectral indices for the early selection of cotton genotypes based on fiber quality traits

Cotton is the world's most important source of natural fiber, a material that begins as a fluffy boll on a plant and ends up as the fabric in our clothes. For the textile industry, the value of this crop depends entirely on the quality of its fibers: how long they are, how strong they are, and how uniformly they hold together. Traditionally, breeders who develop new cotton varieties have had to wait until the plants are fully grown and the bolls are harvested to know if a new variety is any good. They take a sample of the lint, send it to a specialized lab, and run it through a machine that measures these technical properties. This process is accurate but slow, expensive, and requires a large amount of plant material, meaning breeders often cannot test new varieties until they are already quite advanced in their development.

In recent years, a new approach has emerged to speed up this process: using drones to take pictures of crops from the sky. These unmanned aerial vehicles, or UAVs, can capture detailed images of a field in minutes. By analyzing the colors in these images, scientists can calculate numbers that describe how healthy and vigorous the plants are. The idea is that the way a plant looks from above—its color and brightness—might tell us something about the quality of the fiber it will produce inside the boll, long before the harvest. If this connection is strong, breeders could screen thousands of plants early in the season, picking only the best ones to grow further, saving time and money.

A team of researchers in Brazil set out to test whether this idea works for cotton fiber quality. They worked with nineteen different commercial cotton varieties, planting them in a field at the Federal University of Uberlândia. As the plants grew, the researchers waited until the fields were in full bloom, a stage where the plants are covered in open flowers. At this specific moment, they flew a drone over the field to capture high-resolution red, green, and blue images. From these pictures, they calculated four different color-based numbers, known as spectral indices, which act as a summary of the plant's greenness and color saturation. Later, after the cotton was harvested, they took samples from each variety and measured their fiber quality using the standard laboratory machine, which determined traits like fiber length, strength, and a composite score called the Spinning Consistency Index that predicts how well the fiber will spin into yarn.

The researchers found that there was indeed a meaningful link between what the drone saw and what the lab measured. They discovered that two of the color-based numbers were particularly useful. One, which measured the greenness of the leaves, showed a strong connection to the Spinning Consistency Index. Another, which compared the amount of green light to red light, was closely tied to how uniform the fiber length was. In simple terms, the plants that looked greener and more vibrant from the drone's perspective tended to produce fibers that were better suited for high-quality spinning. The team used statistical models to see if they could predict the fiber quality just by looking at these drone numbers. The models were able to explain roughly one-third of the differences they saw between the varieties. While this is not a perfect prediction, it is a significant step forward, suggesting that a drone image taken early in the season can offer a reliable hint about the final quality of the cotton.

However, the study also made it clear that this technology is not a replacement for the traditional lab tests. The drone images could not predict every detail of the fiber quality, and some varieties that looked similar from the sky turned out to have different fiber characteristics. The researchers noted that the best-performing variety in their study, a type called DP 2077, had excellent fiber quality but only showed average color scores in the drone images. This tells us that while the drone can spot general trends, it cannot yet see the full picture on its own. The color of the leaves is influenced by many factors, including how much sunlight the plant gets and how much water it has, which can sometimes mask the genetic potential for fiber quality.

The value of this work lies in its potential to act as a filter. Breeders often have to manage thousands of plants, and testing every single one in a lab is too costly and slow. If a drone can fly over a field and quickly identify the plants that are likely to have poor fiber quality, those plants can be removed from the breeding program early. This allows the breeders to focus their resources on the most promising candidates. The researchers concluded that while the drone-derived numbers are not a magic solution that eliminates the need for lab work, they are a powerful tool to be used alongside it. By combining the speed of aerial imaging with the precision of traditional testing, cotton breeders can make smarter decisions faster, potentially bringing better cotton varieties to farmers and textile mills sooner. The study suggests that as these models are refined with more data from different environments and more diverse plant types, their ability to predict the future quality of a cotton crop will only improve.