| 摘要: |
| 将遥感变化检测技术结合多期遥感影像进行农作物提取，及时、准确地获取作物播种面积信息，对于制定国家/区域农业经济发展规划、指导种植业结构调整，提高农业生产管理水平具有重要的意义。文章根据变化检测方法特点出发，从硬变化检测方法（HCD）和软检测方法（SCD）等2个方面入手，以近几十年国内外重要成果为基础，首先，总述了各自的应用领域；接着分别对HCD和SCD各自的方法分类，方法特点、方法适用性进行了总结。分析结果认为：硬变化检测方法（HCD），侧重于硬变化区域的信息提取，其排他性的二值结果（0， 1）对于中、低分辨率影像，尤其是在作物破碎种植区的识别结果误差较大； 而软变化检测方法（SCD），在混合变化像元区和过渡变化像元区表达出更加丰富的信息，但由于光谱的不稳定性因素，带来类内光谱不稳定性，导致硬变化区域（完全发生变化的区域）的识别造成误差。文章最后结合时空尺度上土地覆盖变化的真实情况，将软、硬相结合的变化检测技术应用前景和发展进行了阐述。 |
| 关键词: 农作物 遥感 软变化检测 硬变化检测 软硬变化检测 |
| DOI：10.7621/cjarrp.1005-9121.20150726 |
| 分类号: |
| 基金项目:北京市教育委员会科技创新平台项目（PXM2013-014225-000051）;北京工业职业技术学院一般课题（bgzyky201518）；国家自然科学基金(41301444);北京市高等学校"青年英才计划"; 高分辨率对地观测系统专项重大科技工程 |
|
| A REVIEW ON CROP MAPPING USING REMOTE SENSING CHANGE DETECTION APPROACH |
|
Zhu Shuang1, Zhang Jinshui2
|
|
1.Beijing Polytechnic College, Beijing 100042, China/ College of Resources Science and Technology/State Key Laboratory of Earth Surface Processes
and Resource Ecology, Beijing Normal University, Beijing 100875, China;2.College of Resources Science and Technology/State Key Laboratory of Earth Surface Processes
and Resource Ecology, Beijing Normal University, Beijing 100875, China
|
| Abstract: |
| It is important to timely and accurately gain cropping areas for agricultural planning, planting structure adjustment and land use management. Remote sensing has played a significant role in crop acreage detection because of its large coverage and short-period detection. Crops undergo a rapid land use/cover change within a short period, and each crop has its own phonological characteristics, which is quite different from the sequence change of natural vegetation. Therefore, dynamic change detection with multi-temporal images is suitable for crop mapping. The methods of crop mapping by change detection can be summarized as two types: hard change detection method (HCD) and soft change detection method (SCD). HCD is the conventional method and has the advantage of accurately identifying crop by phonological growth characteristics. The classification results from HCD are expressed by abundance value ［0,1］. 0 indicates no change in class and 1 indicates change in class. Hence, HCD is not suitable in the fragmental planting areas where mixed pixels are predominant. The results from SCD are between 0 to 1, and have more informative than that from HCD. SCD is mainly applied to express the land cover change characteristics of a specified target land and can identify weak land cover change. However, the spectral variability within a class leads to identification errors in some changed areas. The change status of real earth surface can be divided into three parts: changed area, partially changed area, and unchanged area, according to the continuous changed spectral characteristics from medium-resolution and low-resolution remote sensing images. Discrete and continuous changes are usually coexisting. Discrete region refers to the area that is totally changed from one class (e.g. crop) to other classes (e.g. non-crop), whereas the continuously changed region refers to the area where class type is partially changed. Using either HCD or SCD separately may lead to the identification errors. Therefore, combining the advantage of conventional HCD and SCD, this paper proposed to use soft and hard change detection method (SHCD) for crop distribution mapping. SHCD can identify crop area by the status of land cover change in discretely changed area, and meanwhile estimate crop abundance by the status of crop change and change magnitude in continuously changed areas. Thus, SHCD can improve the accuracy of crop identification. Finally, the advantages and disadvantages of SHCD were also discussed in this paper. |
| Key words: crop remote sensing soft change detection method hard change detection method soft and hard change detection method |
