|
| 光谱与纹理信息结合的黄河三角洲土壤盐渍化信息提取——以垦利区为例 |
| Extraction of soil salinization information by combining spectral and texture data in the Yellow River Delta: A case study in Kenli District, Shandong Province |
| 投稿时间：2021-01-12 |
| DOI：10.13254/j.jare.2021.0025 |
| 中文关键词: 黄河三角洲，土壤盐渍化，光谱特征，纹理特征 |
| 英文关键词: Yellow River Delta, soil salinization, spectral signature, textural feature |
| 基金项目:国家自然科学基金项目（41877003）；山东省重大科技创新工程项目（2019JZZY010724）；山东省"双一流"奖补资金（SYL2017XTTD02） |
| 作者 | 单位 | E-mail | | 黄静 | 山东农业大学资源与环境学院, 土肥资源高效利用国家工程实验室, 山东 泰安 271018 | | | 赵庚星 | 山东农业大学资源与环境学院, 土肥资源高效利用国家工程实验室, 山东 泰安 271018 | zhaogx@sdau.edu.cn | | 奚雪 | 山东农业大学资源与环境学院, 土肥资源高效利用国家工程实验室, 山东 泰安 271018 | | | 崔昆 | 山东农业大学资源与环境学院, 土肥资源高效利用国家工程实验室, 山东 泰安 271018 | | | 高鹏 | 山东农业大学资源与环境学院, 土肥资源高效利用国家工程实验室, 山东 泰安 271018 | |
|
| 摘要点击次数: 884 |
| 全文下载次数: 989 |
| 中文摘要: |
| 土壤盐渍化是制约黄河三角洲农业发展的关键问题，及时准确地掌握土壤盐渍化信息对土地资源保护和开发利用具有积极意义。本研究以黄河三角洲核心区域垦利区2019年4月17日的Sentinel-2遥感影像为数据源，在ENVI和e Cognition软件支持下，利用灰度共生矩阵法提取遥感影像的二阶矩、对比度、熵、相关性等纹理特征信息，结合归一化植被指数（NDVI）、盐分指数（SI）等光谱特征信息，通过预设分类规则实现对黄河三角洲垦利区的盐渍土分类。结果表明，加入二阶矩、对比度、熵、相关性4个纹理特征统计量，再结合光谱信息对垦利区盐渍土进行分类，总体分类精度为92.4%，Kappa系数为0.89，相较于仅利用光谱信息的分类方法，总分类精度提高了10.5个百分点；各分类类别的生产者精度与使用者精度较仅依靠光谱信息分类的分类结果均明显提高，其中中度盐渍土的分类效果最好，其生产者精度与使用者精度最高，分别为95.0%、95.9%。本研究提出利用遥感光谱结合纹理特征实现滨海区盐渍土信息的提取方法，提高了盐渍土分类精度，为准确掌握研究区土壤盐渍化信息提供了新途径。 |
| 英文摘要: |
| Soil salinization is the key problem that restricts the agricultural development in the Yellow River Delta. It is significant to grasp the soil salinization information accurately for the protection and development of land resources. In this study, the Sentinel-2 remote sensing image on April 17th, 2019 in Kenli District, which is the core region of the Yellow River Delta, was used as the data source. Under the softwares ENVI and e Cognition, the GLCM method was used to extract the texture feature information of remote sensing images, such as Second Moment, Contrast, Entropy, and Correlation. Combined with the spectral feature information such as NDVI and SI, the classification of saline soil in the reclamation area was identified by preset classification rules. The results showed that four texture feature statistics of Second Moment, Contrast, Entropy, and Correlation were added, and the spectral information was combined to classify the saline soil in Kenli District. The overall classification accuracy and Kappa coefficient were 92.4% and 0.89, respectively. Compared with the classification method using only spectral information, the overall classification accuracy was improved by 10.5 percent points. The producer and user precision of each classification category were significantly improved compared with the classification results based only on spectral information. The classification effect of moderate saline soil was the best, and the producer and user precision were the highest, which were 95.0% and 95.9%, respectively. In this study, a method of extracting salinized soil information in coastal areas using remote sensing spectrum combined with texture features was proposed, which improved the classification accuracy of salinized soil and provided a new way for accurately grasping soil salinization information. |
| HTML
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
