基于多时相遥感影像的采煤塌陷区典型扰动轨迹识别——以山东省济宁市典型高潜水位矿区为例

本文以济宁典型采煤塌陷地区为研究区域，首先基于1985~2018年Landsat系列遥感影像，利用动态时间规整算法获取本区域典型扰动轨迹特征，然后利用Python决策树分类算法对区域内像元扰动类型和扰动时间进行聚类和识别，从而揭示采矿活动对本区域土地的扰动类型和扰动时间的数量结构、空间分布特征。结果表明：1985~2018年间，研究区范围内存在扰动的像元数量为48517个，面积约4367 km2，占研究区总面积的62.30%，其中扰动最多的年份主要集中在1994~2007年，累计扰动的像元量为35724个，占所有扰动像元的73.78%。1998年扰动像元数量在监测期内最多，像元数为8868个，占所有扰动像元的18.28%。整个研究区域扰动像元数量呈逐步减少趋势，说明该区域整体扰动强度正在减弱，生态环境状况正逐步恢复。本研究探索了多时相遥感地表扰动监测的新方法，研究结果可为区域内生态环境治理提供一定科学依据。

Identification of typical disturbance trajectory in coal mining subsidence area based on multi- temporal remote sensing images

This paper selects the typical coal mining subsidence area in Jining city as the research area, and uses the Dynamic Time Warping algorithm to obtain the typical disturbance trajectory characteristics of the region based on Landsat remote sensing images from 1985~2018. The Python decision making classification algorithm was used to cluster and identify the disturbance type and disturbance time of the region, which reveals the quantitative structure and spatial distribution characteristics of the disturbance type and disturbance time of the mining activity. The results show that between 1985 and 2018, the number of pixels in the study area is 48517, and the area is about 43. 67 km 2, accounting for 62. 30% of the total area of the study area. The most disturbed years are mainly concentrated in 1994~2007. The cumulative number of disturbed pixels is 35724, accounting for 73. 78% of all disturbing pixels. In 1998, the number of disturbed pixels was the most during the monitoring period, and the number of pixels was 8868, accounting for 18. 28% of all disturbing pixels. The number of disturbed pixels in the whole research area is gradually decreasing, indicating that the overall disturbance intensity is weakening and the ecological environment is gradually recovering. This study explores a new method for multi temporal remote sensing surface disturbance monitoring. The research results can provide a scientific basis for region ecological environment management.