Edge Computing-Based Monitoring of Mosquito Breeding Grounds Using the Flatten Method and Neural Network Classifier

The increasing prevalence of dengue hemorrhagic fever in densely populated urban areas of Indonesia highlights the urgent need for effective mosquito breeding site monitoring. Traditional methods, such as manual inspection and larvicide application, are often limited by resource constraints and low coverage. To overcome these challenges, this study presents an edge computing-based solution that leverages the flatten method and a neural network classifier to detect high-risk mosquito breeding conditions in real time. The proposed method achieved strong performance, with a classification accuracy of 96% and an estimated F1 score of 0.95. It also demonstrated excellent efficiency, with an inference time of 1 ms, peak RAM usage of only 1.7 KB, and a flash memory footprint of 15.5 KB. These results affirm the effectiveness of combining the flatten method and neural network classification in an edge computing framework, offering a reliable and scalable approach for autonomous mosquito breeding ground monitoring in urban public health applications.