Urban tree classification using discrete-return LiDAR and an object-level local binary pattern algorithm

Abstract

Urban trees have the potential to mitigate some of the harm brought about by rapid urbanization and population growth, as well as serious environmental degradation (e.g. soil erosion, carbon pollution and species extirpation), in cities. This paper presents a novel urban tree extraction modelling approach that uses discrete laser scanning point clouds and object-based textural analysis to (1) develop a model characterised by four sub-models, including (a) height-based split segmentation, (b) feature extraction, (c) texture analysis and (d) classification, and (2) apply this model to classify urban trees. The canopy height model is integrated with the object-level local binary pattern algorithm (LBP) to achieve high classification accuracy. The results of each sub-model reveal that the classification of urban trees based on the height at 47.14 (high) and 2.12 m (low), respectively, while based on crown widths were highest and lowest at 22.5 and 2.55 m, respectively. Results also indicate that the proposed algorithm of urban tree modelling is effective for practical use.