基于遥感的撂荒耕地提取方法比较研究

Comparative Study on Extraction of Abandoned Farmland Based on Remote Sensing

粮食安全是国家安全的重要基础.全球气候持续变暖引起的粮食减产以及快速镇化背景下农村耕地撂荒现象的不断加剧,严重威胁我国粮食安全.开展撂荒耕地的遥感监测,准确掌握撂荒耕地的空间分布,对于科学制定农业发展措施具有十分重要的意义.本文以我国贵州清镇市为研究区,在分析该地区农作物物候特征以及撂荒耕地光谱特征基础上,利用哨兵2号高分辨率卫星遥感数据,采用单时相波段特征法、植被指数变化检测法和CART决策树分类法分别提取撂荒耕地,并对不同方法进行对比.结果表明:(1)由于季节性作物种植和生长的差异,撂荒耕地与种植耕地不同季节的光谱特征差异明显;(2)基于不同地物光谱特征值差异的单时相波段法容易受到其他光谱值相近的地物的影响,撂荒耕地的提取精度仅有67%,显著低于考虑季相变化特征并结合波段光谱差异的植被指数变化检测法和CART决策树分类法,其提取精度分别达到82%和73%.

Food security is an important foundation for national security.As global climate continues to warm,the resulting reduction in food production poses a greater threat to global food security.At the same time,In the context of rapid urbanization,a large number of rural arable landswere abandoned and the situation becomes more and moreIntensified,which will threaten to our food security seriously.The research of the monitoring of abandoned farmland based on remote sensing will be useful to accurately grasp the pattern of the abandoned farmland and be of great significance to theformulation of agriculturaldevelopment measures.The Qingzhen City of Guizhou Province was chosen as the study area in this paper,the sentinel-2 remote sensing datawas used to extract the distribution of abandoned farmland by the methods of single-temporal band feature detection,thevegetation index transformation detection and CART decision tree classification.The results show that:(1)the spectral features of the abandoned farmland differ significantly in different seasons due to the differences in seasonal crop cultivation;(2)The overall accuracy of single-temporal band feature detection method was only 67%,which was significantly lower than the methods of the vegetation index transformation detection(82%)and CART decision tree classification(73%).