基于合成孔径雷达数据的岩体识别研究——以黑龙江亚布力地区为例

ROCK MASS IDENTIFICATION BASED ON SYNTHETIC APERTURE RADAR DATA: A Case Study of Yabuli Area in Heilongjiang Province

摘要: 在东北植被覆盖区选取试验区，研究不同类型雷达数据对岩体的解译能力和可解译程度，结果显示，高分辨率全极化雷达数据经过极化分解后，对岩体类型的解译效果更佳，但是对其他表征特征属性可解译程度有限.以此为基础，提出一种基于雷达数据和LSTM网络模型的岩体类型识别模型，经评价，预测结果图整体区域分布正确，总体精度评价指标达到83.0%，能够较好地满足地质工作者快速、精准获取岩体信息的需求.

Abstract: The interpretation ability and interpretable degree of different types of radar data to rock mass are studied in a test site of vegetation-covered area in Northeast China. The results show that the high-resolution full-polarization radar data has better interpretation of rock mass types after polarization decomposition, but the interpretable degree of other characteristic attributes is limited. On this basis, the paper proposes a rock mass identification model based on radar data and long short term memory(LSTM) network model. The whole region distribution of prediction result map is correct through evaluation, with the overall accuracy evaluation index reaching 83.0%, which can well help geologists obtain rock mass information quickly and accurately.