Optimized Dual Temporal Gated Multi-Graph Convolution Network for Land Use and Land Cover Classification Incorporating Temporal Feature Tracking and High Resolution Satellite Imagery Analysis

In the invisible world of modern technology, tiny computers known as sensor nodes are scattered across forests, cities, and factories to watch over our environment. These devices, which form what experts call a wireless sensor network, act as the eyes and ears of the digital age, tracking everything from temperature changes to air quality. However, keeping this vast, distributed network alive and secure is a constant struggle. The devices run on limited batteries, and because they often operate in open, unguarded spaces, they are vulnerable to theft, sabotage, and data theft. If the network runs out of power too quickly or if a bad actor slips in to steal information, the entire system can fail. The challenge for scientists is to design a way for these sensors to talk to each other efficiently, organize themselves into groups to save energy, and verify that every message comes from a trusted source, all without draining their batteries or slowing down the flow of information.

A team of researchers has proposed a new, multi-layered system designed to solve these exact problems for networks that rely on edge computing, where data is processed closer to the source rather than in a distant cloud. Their approach, which they call a secure multipath routing system, acts like a highly organized traffic control center for these tiny devices. Instead of letting sensors wander randomly or transmit data in a chaotic stream, the system first builds a structured map of the network using a specialized neural network, a type of computer brain designed to recognize patterns. This initial step organizes the sensors using a Spatial Bayesian Neural Network (SBNN) based hexagonal architecture, which simplifies the task of managing thousands of individual devices and reduces the confusion that often leads to wasted energy.

Once the network is built, the system immediately puts a rigorous security check in place. Before any sensor is allowed to join the conversation, it must prove its identity using a sophisticated encryption method that hides its location and identity even while it searches for data. This process ensures that no fraudulent or unauthorized devices can sneak into the network to steal information or disrupt operations. With the network secured, the system then moves to the critical task of organization. It groups the sensors into clusters, similar to how a school might group students into teams, but it does so dynamically. A bio-inspired algorithm, which mimics the hunting and social behaviors of Humboldt squid, constantly evaluates which sensor is best suited to lead a group. This leader, known as a cluster head, is chosen based on how much battery power it has left, how far it is from other devices, how trustworthy it is, and how quickly it can send data. By constantly re-evaluating these factors, the system ensures that the most capable sensors lead the groups, preventing weaker devices from being overworked and dying prematurely.

To further stretch the life of the network, the system puts the sensors to sleep when they are not needed. Using an advanced learning technique, the network assigns specific time slots to each device, telling them exactly when to wake up, send data, and go back to sleep. This duty cycling prevents the sensors from transmitting constantly, which would drain their batteries in hours. Instead, they operate in a coordinated rhythm, ensuring that data flows smoothly without collisions or unnecessary energy use. When it comes time to send the collected information to a central station, the system does not rely on a single path. Instead, it uses a Triple-Memristor Hopfield Neural Network (TMHNN) to find multiple safe routes for the data. This network learns to avoid dangerous paths or nodes that might be compromised, choosing the most efficient and secure route available. If one path becomes blocked or unsafe, the system instantly switches to a backup route, ensuring that the data always reaches its destination.

The researchers tested this entire system in a simulated environment to see how it performed against other existing methods. The results showed a significant improvement in how long the network could last and how well it delivered data. In their simulations, the new system used about 14 to 21 percent less energy than previous methods, meaning the sensors could operate for a much longer time before needing a battery replacement. It also extended the total life of the network by nearly 20 to 27 percent, keeping the system active for more rounds of data collection. Perhaps most importantly, the system delivered a higher percentage of data packets successfully, with improvements ranging from 18 to 27 percent compared to older techniques. By combining secure identity checks, smart grouping, energy-saving sleep schedules, and multiple secure routes, this new approach offers a robust way to keep wireless sensor networks running efficiently and safely, ensuring that the data we rely on from the natural world remains secure and available for as long as possible.