基于粒度计算的高分辨率遥感影像耕地信息提取研究

Remote sensing technology has become one of the important scientific and technological means of obtaining quantity and spatial distribution information of cultivated land resources. Cultivated land in southern China is scattered, small-sized, and of obvious regional and seasonal differences, which make the extraction of cultivated land information from high-resolution remote sensing image affected by many uncertainties such as small spectrum divisibility, complex geometry, varied texture pattern, and polytropical seasonal features etc. . Since granular computing can effectively deal with uncertain information by structured thinking and methods, its theory and method are used in this project for research on the extraction of cultivated land information from high-resolution remote sensing image. Including: (1) to obtain typical features of cultivated land based on the scale space of high-resolution remote sensing images; (2) to determine the best scale of cultivated land information extraction according to the scale effect of the cultivated land information; (3) to achieve intelligent extraction of cultivated land information combining with granular computing theory and object-based principle. The project intends to establish an intelligent method for extraction of cultivated land information from high-resolution remote sensing image, with which reliable basic data will be available for the implementation of precision agriculture in China's highly decentralized countryside, and on the other hand updated and accurate information will be provided for the fulfillment of cultivated land protection and national geographic conditions monitoring.

遥感技术已成为获取耕地资源数量与空间分布信息的重要科技手段之一。我国南方地区的耕地呈现分布散规模小且区域性和季节性差异明显的特点，这使得高分辨率遥感影像耕地信息的提取容易受到光谱可分性小、几何形态复杂、纹理模式多样、季相特征多变等诸多不确定因素的影响。粒度计算能够利用结构化思想和方法有效地处理不确定性问题。为此，本项目利用粒度计算的理论与方法开展高分辨率遥感影像耕地信息提取研究，具体包括：（1）基于高分辨率遥感影像尺度空间，研究耕地典型特征的获取方法；（2）依据耕地信息的尺度效应，研究耕地信息最佳提取尺度的选择方法；（3）结合粒度计算理论和面向对象思想，研究耕地信息的智能提取方法。本项目拟通过深入研究高分辨率遥感影像耕地信息的智能提取方法，一方面为我国农村高度分散状况下精细农业的实施提供可靠基础数据，另一方面为我国耕地资源保护与地理国情监测的落实提供及时准确的信息。

耕地作为人类长期赖以生存的自然资源，是国家粮食安全的重要物质保障。随着我国地理国情监测政策在全国范围的逐步推广实施，急需发展快速高效的耕地提取和监测方法。高分辨率遥感影像包含丰富而详尽的地物信息，它可以准确反映出耕地的空间分布信息。本项目主要完成了以下方面的研究：高分辨率遥感影像分割方法，高分辨率遥感影像分类方法，以及高分辨率遥感影像耕地提取方法。在国内外遥感领域的一级期刊发表或录用论文7篇，其中EI/SCI论文5篇，CSCD论文2篇。所提方法能将高分辨率遥感影像耕地信息提取的精度提高到80%左右，一定程度上缓解了耕地提取受到光谱可分性小、几何形态复杂、纹理模式多样、季相特征多变等因素的影响。本项目的实施可为我国耕地资源保护与地理国情监测提供重要的基础信息参考。