Research on the optimization path of geographic computing efficiency in large-scale vegetation dynamic monitoring

Aiming at the problem of low geographic computing efficiency in large-scale vegetation dynamic monitoring, this paper proposes a parallel computing algorithm (ABAPA) based on adaptive blocking and attention mechanism. The algorithm first pre-processes the vegetation remote sensing data, and then realizes adaptive blocking according to the spatial distribution density of vegetation. The key vegetation areas are identified through the attention mechanism and different computing priorities are assigned, and the efficiency is improved by combining the load-balanced parallel scheduling method. The experiment uses 50GB, 100GB, and 200GB of vegetation remote sensing data from different regions (tropical rainforest areas, temperate grassland areas, etc.) and resolutions (10m, 30m, etc.), and compares ABAPA with traditional parallel algorithms and optimized parallel algorithms based on fixed blocking. The results show that when processing 100GB of data, ABAPA reduces the computing time by 40% (from 2800 seconds to 1680 seconds) compared with the traditional algorithm and 25% (from 2240 seconds to 1680 seconds) compared with the fixed block algorithm; the overall monitoring accuracy reaches 95.6%, which is equivalent to the traditional algorithm (95.2%); the memory usage is reduced by 30% (from 120GB to 84GB) compared with the traditional algorithm. The study shows that ABAPA can significantly improve the efficiency of geographic computing while ensuring the monitoring accuracy, and provide an efficient solution for large-scale vegetation dynamic monitoring.