International Journal of Remote Sensing Vol. 33, No. 22, 20 November 2012, 7099–7116 An object-based analysis filtering algorithm for airborne laser scanning MENGLONG YAN*†‡, THOMAS BLASCHKE§, YU LIU‡ and LUN WU‡ †Key Laboratory of Spatial Information Processing and Application System Technology, Chinese Academy of Sciences, Beijing, China ‡Institute of Remote Sensing and Geographical Information Systems, Peking University, Beijing, China §Z_GIS – Centre for Geoinformatics and Department for Geography and Geology, University of Salzburg, A-5020 Salzburg, Austria (Received 14 July 2011; in final form 28 February 2012) Ground filtering is a key process to derive digital terrain models from airborne laser scanning data. Although many methods have been developed to tackle the filter- ing problem, it has not been fully solved so far. Current algorithms mainly focus on neighbourhood-based or directional filtering approaches. A new object-based analysis (OBA) method is proposed in this article. First, a grid index algorithm accelerates access to unorganized cloud points. Then, a segmentation algorithm is deployed based on the index, and objects are obtained. A filtering logic that uti- lizes the objects’ characteristics is designed. Following this, the performance of the method is comprehensively tested using publicly available International Society for Photogrammetry and Remote Sensing (ISPRS) test data sets for nine urban and six rural regions, and the results are compared to those of eight other algorithms. The OBA method implemented in this article reveals good results without scene-wise optimization of the parameters, and it ranks third or fourth in most of the cases. 1. Introduction Airborne laser scanning (ALS) is a technology for the quick acquisition of digital ter- rain models (DTMs) and digital surface models (DSMs) (Ackermann 1999, Sithole and Vosselman 2004). The method is also utilized for extracting features such as buildings (Sohn and Dowman 2007) and vegetation (Wagner et al . 2008). As is well known, ground and non-ground back echoes are confusingly mixed in raw ALS data. When aiming to differentiate topographic information, the basic task of processing ALS data is to distinguish bare ground points from object points. Due to the complexity of terrain, a full automation of the point filtering process is not possible (Sithole and Vosselman 2004). Hyyppä et al . (2004) report from practical tests that laser scanning data even include low points under the ground level. These may be real returns or falsely interpreted elevation values from strong backscatters. Therefore,the typical procedures in existing approaches require a lot of human inter- action, which is usually labour-intensive and time-consuming. Many research articles focus on filtering algorithms (Kraus and Pfeifer 1998, Pfeifer et al . 1999, Axelsson 2000, Vosselman 2000, Roggero 2001, Sithole 2001, Brovelli et al . 2002, Elmqvist 2002, *Corresponding author. Email: yanmenglong@gmail.com International Journal of Remote Sensing ISSN 0143-1161 print/ISSN 1366-5901 online © 2012 Taylor & Francis http://www.tandfonline.com http://dx.doi.org/10.1080/01431161.2012.699694 Downloaded by [Universitat Salzburg] at 07:37 20 July 2012