论文部分内容阅读
建立实用的遥感图像分析系统涉及诸多因素,其中,形成有效的目标提取算法体系和设立适于目前人工智能水平的基于知识的框架是两个核心问题。本文提出了一种利用对光谱特征空间分布进行可视化图形处理的新的分类方案,继而以光谱、纹理和空间知识对分类结果进行优化。这种集可视化分析、栅格GIS处理功能和人类判读专家知识于一体的系统为解决遥感图像地物识别这一复杂问题提供了有效途径。
There are many factors involved in establishing a practical remote sensing image analysis system. Among them, forming an effective target extraction algorithm system and establishing a knowledge-based framework suitable for the current level of artificial intelligence are two core issues. This paper presents a new classification scheme that utilizes the visualization of spatial distribution of spectral features, and then optimizes the classification results based on spectral, texture and spatial knowledge. This set of visual analysis, grid GIS processing functions and human interpretation expert knowledge in one system to solve the complex phenomenon of remote sensing image recognition provides an effective way.