Abstract: Zhejiang province has implemented the “Intelligent Cultivated Land Protection” system, leveraging the province-wide, extensive, and elevated tower resources, combined with "towers+video terminals+AI algorithms" to comprehensively adopt both human and technological surveillance measures. Traditionally, two-dimensional site selection schemes based on experiential judgment have suffered from low accuracy, inefficiency and high subjectivity, while coverage rate and cost-effectiveness are key metrics for evaluating the quality of surveillance network design. In this context, this paper utilizes high-precision 3D real-world data and viewshed analysis algorithms, combined with 3D simulation technology, to construct a visual mapping dataset of candidate surveillance points and cultivated land plots. By studying greedy algorithms, genetic algorithms, and simulated annealing algorithms, this paper proposes a hybrid approach that combines simulated annealing with genetic algorithms, using an improved greedy algorithm to provide high-quality initial solutions for the genetic algorithm. This hybrid genetic algorithm (GA-SA) enhances the accuracy and cost-effectiveness of surveillance site selection. Additionally, a cultivated land plot surveillance visualization system is developed based on the Cesium 3D engine, providing critical visualization support tools for cultivated land management and protection.

Key words: intelligent farmland protection, viewshed, genetic algorithm, site optimization, 3D real scene