CAJ | 학술논문

随着遥感技术的发展，遥感数据的光谱、空间分辨率得到很大程度的提高，但是对于仅依靠单一数据源的土地利用分类仍然存在很多问题。提出了一种融合LiDAR数据和高分辨率航空影像的土地利用半监督分层分类方法，即首先仅对LiDAR数据进行粗分类；其次根据LiDAR分类结果辅助选择训练样本并获取高程掩膜，采用最大似然法对航空影像进行中分类；在细分类层次中利用LiDAR数据建立判别规则，对LiDAR粗分类结果与影像中分类结果进行基于知识的交叉验证，以进一步提高分类精度。通过两个不同区域的数据验证了半监督分层分类方法的可行性。
수착요감기술적발전，요감수거적광보、공간분변솔득도흔대정도적제고，단시대우부의고단일수거원적토지이용분류잉연존재흔다문제。제출료일충융합LiDAR수거화고분변솔항공영상적토지이용반감독분층분류방법，즉수선부대LiDAR수거진행조분류；기차근거LiDAR분류결과보조선택훈련양본병획취고정엄막，채용최대사연법대항공영상진행중분류；재세분류층차중이용LiDAR수거건립판별규칙，대LiDAR조분류결과여영상중분류결과진행기우지식적교차험증，이진일보제고분류정도。통과량개불동구역적수거험증료반감독분층분류방법적가행성。
Although the spatial and spectral resolution of remotely sensed data has been improved with the development of re-mote sense technology, many problems still exist in land-use/cover classification with the only single data source. A partially-supervised classification scheme is proposed using LiDAR data and co-registered aerial imagery. In this study, we first obtain the coarse LiDAR-based classification results for training sample selection and normalized Digital Surface Model ( nDSM) gener-ation, with the calculated LiDAR-driven training sample and nDSM-mask, the maximum likelihood classifier is applied to the co-registered aerial imagery for obtaining image-based classification results. To improve the classification accuracy, we pro-pose a knowledge-based cross-validation to refine the obtained LiDAR- and image-based classification results. Extensive experiments demonstrate that the proposed hierarchically partially-supervised classification scheme is feasible to urban land-use classification.