Automated Detection of Small Water Accumulations in Aerial Images Acquired by Drones: A Genetic Algorithm-Based Method

Resumo

The use of drones in combating mosquito breeding sites, such as Aedes aegypti, is already a consolidated reality. However, although the presence of stagnant water is an essential condition for the reproduction of these vectors, computer vision approaches reported in the literature rarely address the direct detection of water in suspected objects and scenarios, representing a significant research gap. In this work, we propose a method based on Genetic Algorithms (GA) for creating a water index, obtained from arithmetic combinations of the image spectral bands. Each GA chromosome, when decoded, generates an index that is evaluated by a fitness function aimed at minimizing the difference between the pixel values produced by the application of the index and the expected values, extracted from manually annotated images. In summary, at the end of the evolutionary cycle, the GA produces a water index that, when applied to an image captured by the drone, generates a binary image highlighting the regions corresponding to the presence of water. The results obtained (accuracy = 87.2%, precision = 85.2%, and sensitivity = 95.2%), demonstrate the feasibility of the proposed method for detecting small water accumulations in objects and scenarios potentially associated with breeding sites, thus enhancing the effectiveness of drones in surveillance and epidemiological control activities.