| 摘要: |
| 采用高分辨遥感影像提取农作物信息在北方区域得到了深入的研究和广泛应用，但受到南方地形复杂、气候差异等条件的影响，采用遥感技术提取南方农作物信息的研究甚少。鉴于此，该文以广西鹿寨县为例，采用2012年9月的高分辨率RapidEye遥感影像为主要研究数据源，利用面向对象的遥感图像分析方法，结合高程数据模型DEM和GPS野外采集数据，综合桑树在遥感影像上的形状特征、颜色特征、光谱特征、纹理特征、位置特征（拓扑特征）提取桑树信息，并对其分类结果进行精度验证和评价。研究结果表明，提取桑树信息的总精度达到9196%，Kappa系数达到08391，满足使用的精度要求。面向对象的图像分析方法为高分辨率遥感图像信息提取农作物信息提供了新的思路，该方法与基于像元的分类技术相比，突破了过多地关注地物局部细节而难以提取地物的整体结构信息的缺陷，避免了“椒盐现象”的产生，有效地提高了农作物识别的精度和效率，这对提取广西大宗农作物信息具有一定的借鉴意义。 |
| 关键词: 高分辨率 面向对象 RapidEye影像 桑树 信息提取 广西鹿寨 |
| DOI：10.7621/cjarrp.1005-9121.20150207 |
| 分类号: |
| 基金项目:教育部科学技术研究重点项目（2010168） |
|
| INFORMATION EXTRACTION OF SILKWORM-CROPPING AREAS BASED ONHIGH SPATIAL RESOLUTION REMOTE SENSING IMAGES——A CASE STUDY OF LUZAI COUNTY, GUANGXI, CHINA |
|
Xie Guoxue1, Huang Wenxiao2, Lu Yuan1, Yang Rujun3
|
|
1.School of Geography and Planning, Guangxi Teachers Education University, Nanning 530001, China/Key Laboratory of Beibu Gulf Environmental Evolution and Resources Utilization, Ministry of Education, Nanning 530001, China;2.Office of Guangxi Agriculture Regional Planning Committee, Nanning 530022, China;3.Land and Resources Information Center of Guangxi, Nanning 530028, China
|
| Abstract: |
| Using high-resolution remote sensing image to extract crop information has been deeply studied and widely used in the northern region. However, there was little research in the south due to the influence of complex terrain, climatic differences and other conditions. The main purpose of the article was to extract silkworm corps information combining the specific situation with referring predecessor' research results. Taking Luzhai County in Guangxi as an example, and utilizing the RapidEye high resolution remote sensing images in September 2012 as the main source of research data, using the methods of the object-oriented image analysis combining with the elevation data model DEM and GPS field acquisition interpretation and field crops data, this paper extracted silkworm crops information based on the comprehensive information of its shape features, color characteristics, spectral characteristics, texture characteristics and location characteristics (topological features) reflected on the remote sensing images. Finally, the classification results will be accurately verified and evaluated. The results showed the overall accuracy of silkworm crop information extracted was up to 9196%, and the Kappa coefficient reached 08391. As a result, the validation and evaluation results met the precision requirement. It also showed that the method applied to extract the southern crops information was really practical. Finally, it concluded that the object-oriented remote sensing image analysis method and technology provided a good new way for the high-resolution remote sensing image feature information extraction. Compared with the remote sensing image method of pixel-based classification technique, this method not only broken through the shortcomings of difficulty in extracting entire structural information of the culture features due to paying too much attention to partial details, but also avoid the "salt and pepper phenomenon ". What is more, it can effectively improve the accuracy and efficiency of crop information identification. The object-oriented remote sensing image classification method provided a better solution to solve the problems including land broken, mixed planting crops and difficulty to extract crop information in southern region. With the continuous improvement of image resolution, the method will gradually replace the remote sensing image method of pixel-based classification, which will become the hotspot in the research field. |
| Key words: high resolution object-oriented RapidEye image silkworm crop information extraction |
