In traditional agriculture, crop disease control often relies on the extensive spraying of chemical pesticides. However, this “one-size-fits-all” approach not only wastes resources but also risks environmental pollution and threats to human health. With global climate change and increasing pathogen resistance, reducing pesticide dependence through smarter, more precise methods while ensuring crop yields has become a major challenge for sustainable agriculture. Is there a technology that can rapidly identify diseases and apply pesticides precisely, achieving “targeted treatment”?
Dr. Roaf Ahmad Parray from ICAR-indian agricultural research institute (ICAR-IARI) and his colleagues provide an answer in a study published in Frontiers of Agricultural Science and Engineering (DOI: 10.15302/J-FASE-2024572). In this research, an international team of scientists from India, Denmark, and the United States developed an innovative technology integrating spectral sensors, machine learning models, and an intelligent spraying system, successfully applying it to control black rot disease in cauliflower. This technology, comprising three core components—non-destructive detection, intelligent decision-making, and targeted pesticide application—significantly reduces pesticide use and offers new insights for green agriculture.
Traditional disease detection relies on manual observation, which is time-consuming, labor-intensive, and error-prone. The researchers adopted a novel approach, using spectral sensors to capture reflected light signals from plant leaves. Healthy leaves and those infected with black rot exhibit distinct differences in light reflection, particularly in visible and near-infrared bands. By analyzing these “spectral fingerprints”, the sensors rapidly assess plant health. However, how can disease-infected areas be accurately distinguished from healthy ones within vast spectral data? To address this, the researchers introduced machine learning algorithms, comparing decision trees and support vector machines (SVM). Results showed that the SVM model outperformed the decision tree, achieving a test accuracy of 96.7% versus 89.9%. This high-precision model was embedded into the intelligent spraying system’s control unit, serving as the “brain” to determine pesticide application.
Traditional spraying equipment often covers the entire farmland, while the intelligent spraying system in this study is as precise as a “surgeon”. When the sensor detects a diseased area, the system sprays pesticides only on the diseased part through a micro pump and a special nozzle; if a plant is identified as healthy, the spraying function is automatically turned off. Field trials have shown that this system successfully identified and treated 75% of the diseased plants in a 100-square-meter cauliflower test field, and avoided mis-spraying 87.5% of the healthy plants. Compared with traditional backpack sprayers, the intelligent system reduced the amount of pesticides used by 72.5% and saved 21% of the spraying time.
At the same time, the equipment uses low-cost materials and open-source hardware to ensure that small-scale farmers can afford it. The distance between the sensor and the spraying unit of the equipment has been optimized and can work stably within the range of 25–45 centimeters, adapting to farmlands with different planting densities. In addition, the system only needs to regularly calibrate the whiteboard reference value, and it is simple to operate, making it suitable for farmers lacking professional skills.
Currently, the researchers have completed the preliminary verification in the experimental field of the Indian Agricultural Research Institute. Future research will explore the applicability of this technology to other crops (such as tomatoes, potatoes) and diseases (such as downy mildew), and expand its application by combining with drone technology.
DOI: 10.15302/J-FASE-2024572