摘要
针对农作物对地抽样调查工作中村级抽样单元野外调查工作量大的问题,研究了一种基于高分数据的村级农作物种植面积提取方法。以辽宁省北镇市孟家村SPOT5遥感影像为实验数据(主要包括农田、裸地、道路、大棚、水体、房屋等类型),基于eCognition平台,根据局部方差法筛选地物最优分割尺度,按照最优分割尺度从大到小组织实验区地物分割次序,结合影像对象的光谱、形状等特征,建立分类规则,完成了村级农作物种植面积的提取。该方法有效避免了在影像分割时参数反复试错带来的复杂性和随机性,提高了农作物种植面积面向对象分类精度和效率。通过野外调查样地进行精度验证表明,精度达到90.4%,为提高农作物种植面积对地抽样调查工作效率,减少野外调查的工作量提供技术支撑。
In order to solve the hard field investigation work problem in conditional sampling survey, a novel obejct- oriented crop planting area extraction method was proposed through high spatial resolution remote sensing image. The method was tested with SPOT 5 over a study area in Mengji- acun village, Beizhen County, Liaoning Province of China. The main features of the study area included cropland, wasteland, road, greenhouse, water, construction land and trees etc. Using the local variance method based on feature＇s object form segmentation, the optimal parameters for segmentation of the feature were obtained by eCognition. The authors organized the segmentation by these parameters in descending order. After that, the crops were mapped using object - oriented classification method based on the rlues correspond with the features of spectral and shape. High mapping precision of 90.4 % was acquired through field investigation. The method could provide support for improving the efficiency of crop sampling survey work and reducing the workload of field investigation.
出处
《资源开发与市场》
CAS
CSSCI
2014年第5期515-518,F0002,共5页
Resource Development & Market
基金
国家自然科学基金项目(编号:42171516)
教育部人文社会科学研究规划基金项目(编号:12YJA790016)
高等学校博士学科点专项科研基金项目(编号:20113424110002)
