Unmanned Aerial Vehicle (UAV) surveillance has emerged as a transformative tool in disaster management and response. In disaster scenarios, maximizing UAV network coverage is essential for extracting maximum environmental information. A majority of the planar coverage maximization solutions are primarily inapplicable to non-convex and disconnected regions. This paper presents a novel framework that processes input images, extracts boundaries, and reduces their dimensionality using the Ramer–Douglas–Peucker algorithm to generate a union of polygons, which may exhibit non-convexity, disconnectedness, or both. It then implements a genetic algorithm to produce a high-quality sensor configuration. Our approach innovatively integrates Monte Carlo sampling and directly encodes sensor positions into the chromosome, optimizing the use of problem geometry.