基于语义的图像低层可视特征提取及应用

为实现图像低层可视特征提取及其智能语义推理,从遥感图像解译入手,结合灰度共生矩阵和模糊C均值分类器提取图像纹理特征。构造基于灰度形态学的多尺度多结构元素边缘检测算子,提取特征知识。构建基于断层带的多源地学数据语义推理模型。以成都附近的断层为研究对象,进行语义推理验证,其解译结果与专家实地解译情况相符,初步验证该模型的可行性,使图像的机器分析结果更加贴近专业人员的目视解译,为地学研究数字化和遥感图像解译信息化提供参考。

Extraction and Application of Image Low-level Visual Features Based on Semantics

In order to realize extraction of image low-level visual features and semantic reasoning, this paper starts from remote sensing image explanations, combines Gray Level Co-occurrence Matrix（GLCM） and Fuzzy C-Means（FCM） classifier to extract texture feature, then detects edge by multi-scale and multi-structuring elements based on grayscale morphology, finally constructs multi-sources geological data based on the fault zone and uses the Chengdu parcels to test and verify the model. The results completely coincide with the expert＇s field studies, which demonstrates the feasibility of this model, makes the results of machine analysis closer to results of visual interpretation, and provides valuable preferences fordigitalization of the earth science study and informationization of image interpretation.