Pixel-based and object-oriented classifications of airborne LiDAR and high resolution satellite data for building extraction

9th IGRSM International Conference and Exhibition on Geospatial & Remote Sensing

地球科学;计算机科学

Al-Nahas, F.^1 ; Shafri, H.Z.M.^1

Deparment of Civil Engineering, Faculty of Engineering, Universiti Putra Malaysia (UPM), Serdang

43400, Malaysia^1

关键词: Building extraction;    Classification accuracy;    Classification approach;    High resolution satellite data;    High spatial resolution;    High spatial resolution satellite imagery;    Light detection and ranging;    Object oriented classification;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/169/1/012032/pdf
DOI  :  10.1088/1755-1315/169/1/012032

学科分类：计算机科学（综合）

来源: IOP

【 摘 要 】

Building extraction from high spatial resolution data is a challenging task in the remote sensing community because of the spectral similarity between man-made objects, such as buildings, roads, and parking lots, in urban areas. This study utilizes two data types, Worldview-3 (WV3) and airborne Light Detection and Ranging (LiDAR), to extract buildings. The main goal of this study is to investigate the capability of these data sources and its effectiveness in fusing both to extract buildings. Different classification approaches, including pixel-based and object-oriented (OO) approaches, were applied to single WV3 and WV3 and LiDAR fused data. The support vector machine (SVM) was used for both classification approaches. Results show that the OO classification accuracies produced from the fused dataset was higher than that of the pixel-based dataset with 96% accuracy. Our findings also demonstrate that the fusion of LiDAR data with high spatial resolution satellite imagery can improve classification accuracy, especially for building extraction. The fusion of LiDAR data can also decrease the effect of spectral similarity between different man-made objectives in urban areas. The results also show that the OO approach has significant potential for building extraction by utilizing the WV3+LiDAR dataset.