부산대학교 도서관

This paper presents a methodology to support farm decision-making by characterizing regional potential and climatic risks involved during agricultural crop cycles. It introduces agricultural zoning based on a system that classifies digital images of agroclimatic indices. The methodology uses as a first step analysis and processing of climatic data from weather stations through GIS tools. These results are interpolated to generate images of several limiting parameters of agricultural crops. In the second step, these images are segmented using techniques, such as regional growing segmentation by pixel aggregation and regional split/merging segmentation. By using the resulting description of characteristics, the system allows linking the region to recorded geodesic data, generating geo-referenced maps for data storage, information overlay, derivative maps generation, and vector analyses of risk factors necessary in each image. The system generates a matrix of results corresponding to a single image of pseudo-regions. These regions are merged into larger similar regions, following theoretical decision methods. As the result, the system makes it possible to generate maps containing homogeneous zones with optimized planting dates.