Kun Liu

Classification of Big Point Cloud Data using Cloud Computing

Kun Liu Jan Boehm

University College London, United Kingdom

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XL-3/W3, 553-557, 2015


Abstract:

Point cloud data plays an significant role in various geospatial applications as it conveys plentiful information which can be used for different types of analysis. Semantic analysis, which is an important one of them, aims to label points as different categories. In machine learning, the problem is called classification. In addition, processing point data is becoming more and more challenging due to the growing data volume. In this paper, we address point data classification in a big data context. The popular cluster computing framework Apache Spark is used through the experiments and the promising results suggests a great potential of Apache Spark for large-scale point data processing.

Paper:

PDF

Results:

Online Result (Google Cardboard)

Images:

rdd

Figure 1: The directed acyclic graph (DAG) in our implementa- tion using Apache Spark.

time

Figure 2: We also explore the scalability of our implementation by exe- cuting the same test using different data sets of varying sizes as shown in the figure.

res

Figure 3: The classification results of point clouds acquired by mobile mapping system.