AgriMind-PestGuard: An Explainable and Blockchain-Secured Pest Management System for Paddy Farming

Rice is the daily bread for billions of people, yet the fields that grow it face a constant, silent war. Tiny insects, from planthoppers to stem borers, can devour a significant portion of the global harvest before a farmer even notices they are there. For generations, the solution has been a mix of careful observation and the heavy hand of chemical sprays. Farmers often spray entire fields to be safe, a practice that is costly, harmful to the environment, and frequently unnecessary because the pests are not actually present. In recent years, scientists have tried to use artificial intelligence to help, teaching computers to recognize pests from photos. However, these digital tools have often felt like black boxes; they give an answer but no reason, leaving farmers unsure if they can trust the machine. Furthermore, the data these systems collect is often stored in central servers where it can be altered or lost, creating a gap in trust between the technology and the people who rely on it.

A team of researchers has now built a new system called AgriMind-PestGuard to bridge these gaps. Their work combines three distinct technologies into a single, cohesive tool designed specifically for paddy farming. First, they created a sophisticated computer vision system that does not just look at an image but understands the specific shapes and patterns of pests across different sizes. Second, they added a layer of transparency that explains why the computer made a decision, pointing out the specific features, like wing patterns or body shapes, that led to the identification. Finally, they secured the entire process using a digital ledger technology that acts as an unchangeable record book, ensuring that every detection and every recommendation is permanently and honestly logged.

The core of this system is a deep learning model trained on a massive collection of 58,400 real-world images taken from twelve different farms. Unlike previous systems that might struggle when a pest is tiny or when it looks very similar to another species, this new model uses a multi-scale approach. It examines the image at five different levels of detail simultaneously, allowing it to spot the fine features of a small insect just as effectively as the broad shape of a larger one. In testing, this approach proved remarkably accurate. The system correctly identified fourteen different major pest species with an accuracy of 97.83 percent. It successfully distinguished between pests that look nearly identical, such as the brown planthopper and the white-backed planthopper, a task that has historically confused both humans and machines.

What sets this system apart from other artificial intelligence tools is its ability to explain itself. In many advanced computer systems, the decision-making process is hidden, making it difficult for a farmer to know if the result is reliable. The researchers solved this by creating a module that generates a visual map highlighting exactly which parts of the insect the computer focused on. More importantly, it translates these visual clues into biological concepts that an entomologist would recognize. For instance, instead of just showing a red spot on a wing, the system can explain that it identified the pest based on the specific pattern of veins on that wing or the shape of its antennae. When experts reviewed these explanations, they agreed with the computer's reasoning 93.7 percent of the time. This level of transparency turns the system from a mysterious oracle into a trusted partner that farmers can understand and verify.

To ensure that the data generated by this system cannot be tampered with, the researchers anchored every detection to a blockchain network. This technology functions like a shared, digital ledger where every entry is locked in a way that makes it impossible to alter past records without everyone noticing. In this system, whenever a pest is detected, the image hash, the location, the time, and the recommended action are recorded instantly. This creates a permanent, unchangeable history of pest activity and pesticide use. The system is fast enough to handle over 1,200 of these records every second, ensuring that even in a busy farming season, no data is lost or delayed. This layer of security builds a chain of trust among farmers, regulators, and researchers, allowing them to rely on the data for making critical decisions.

The true test of the system came during real-world trials across twelve farms in three different climate zones. The results showed that the system did more than just identify pests; it changed how farmers managed their crops. By using the precise recommendations from the system, farmers reduced their use of pesticides by 34.2 percent compared to their usual practices. This reduction was not achieved by guessing, but by targeting only the areas where pests were actually detected. Consequently, the crops suffered less damage, and the farmers saw their rice yields increase by 8.7 percent. The system also lowered the overall cost of farming per hectare, proving that high-tech precision can be both economically and environmentally beneficial.

This work represents a significant step forward in how technology can support agriculture. By combining high-accuracy detection with clear explanations and secure record-keeping, the researchers have created a tool that respects the intelligence of the farmer while leveraging the power of modern computing. The system does not replace the farmer's knowledge but enhances it, providing a reliable, transparent, and secure way to protect one of the world's most important food sources. The success of these trials suggests that when artificial intelligence is designed with human trust and data integrity in mind, it can deliver tangible improvements in the field, turning the complex challenge of pest management into a manageable, precise, and sustainable practice.