Mobility pattern analysis of ship trajectories based on semantic transformation and topic model

Abstract

Recognition and understanding of ship mobility pattern have great significance for intelligent maritime applications, i.e. route discovery and anomaly detection. Besides a number of pattern discovery techniques currently derived from ship trajectory, topic modeling popular in the field of Natural Language Processing may provide a novel way to detect implicit patters underlying massive ship trajectories treated as documents. This paper is motivated to apply a semantic analysis method to explore potential mobility patterns from ship trajectories in inland river by combining semantic transformation and topic model. A coarse-grained semantic transformation model is firstly defined to translate each ship trajectory into a document containing a series of sequential motion words. A motion word is generally a synthetic semantic representation of three trajectory features (location, course and speed). All ship trajectories can then be examined and analyzed as a document corpus. A classic topic model (i.e. Latent Dirichlet Allocation, LDA) is employed to explore hidden ship mobility patterns from trajectory document corpus. The effectiveness of this approach is illustrated through a case study at Wuhan waterway, located at middle stream of Yangtze River in China.