利用空间特征的平原地区灌溉沟渠优化提取  

Optimal extraction method of irrigation ditches in plain areas using spatial characteristics

作　　者：方炫 张同亮 彭明国[1] 蒋圣 张秋亚 FANG Xuan;ZHANG Tongliang;PENG Mingguo;JIANG Sheng;ZHANG Qiuya(School of Urban Construction,Changzhou University,Changzhou 213164,China;Siyang County Water Conservancy Bureau,Suqian 223700,China;School of Mechatronics and Information,Wuxi Vocational Institute of Arts and Technology,Wuxi 214206,China)

机构地区：[1]常州大学城市建设学院,江苏常州213164 [2]泗阳县水利局,江苏宿迁223700 [3]无锡工艺职业技术学院机电与信息工程学院,江苏无锡214206

出　　处：《常州大学学报（自然科学版）》2023年第4期43-51,共9页Journal of Changzhou University：Natural Science Edition

基　　金：国家自然科学基金资助项目(41871313);江苏省自然科学基金面上资助项目(BK20161118)。

摘　　要：针对异物同谱和植被覆盖影响沟渠提取效果的问题,提出一种基于无人机影像和空间特征对提取过程进行优化和补偿的灌溉沟渠提取方法。首先,采用尺度参数估计(Estimation of Scale Parameter,ESP)方法确定最优分割参数;其次,通过光谱均值及标准差构建特征空间,使用最近邻方法获得初始分类结果;最后,利用长宽比、相对边界指标、边界规整度、面积等空间几何和关系特征,对初始分类结果进行优化提取。试验发现,对于空间分辨率为0.07 m的平原灌区影像,最优分割参数为61,基于空间特征的优化提取方法将提取精度提高到92.3%,较优化前提高了7.6%。该方法能有效解决杂草及异物同谱所导致的沟渠缺失及部分断连问题,提高平原地区灌溉沟渠提取的精度。The capability of applying remote sensing data on ditch extraction is reduced because of vegetation coverage and different objects with same spectrum.To solve the problem,an integrated framework based on unmanned aerial vehicle imagery using spatial features to optimize and compensate the extraction process is proposed for irrigation ditch extraction.Firstly,the optimized segmentation parameter was achieved using the estimation of scale parameter algorithm(ESP).Secondly,the feature spaces for classification were constructed by spectral mean and standard deviation,and the initial classification was conducted by using the supervised k-nearest neighborhood classifier.Finally,the spatial geometric and relational features including aspect ratio,relative boundary index,boundary regularity and area,were used to optimize the initial classification results.The result shows that the optimal segmentation parameter is 61 for the plain irrigation area with 0.07 m resolution,and the optimization extraction method based on spatial features improves the extraction accuracy to 92.3%,which is 7.6%higher than that before optimization.The proposed method can effectively solve the problems of ditch missing and partial disconnection caused by the same spectrum of weeds and different objects with same spectrum,improving the accuracy of ditch extraction in plain irrigation areas.

关 键 词：灌溉沟渠 面向对象法 无人机影像 空间特征 最近邻分类 

分 类 号：X87[环境科学与工程—环境工程]