A behavior-informed adaptive algorithm for hierarchical compression of ship trajectories

Abstract

The Automatic Identification System (AIS) plays a pivotal role in maritime monitoring, yet its high-frequency data often cause redundancy, affecting storage and downstream analysis. Existing compression algorithms often fail to capture vessel behavior semantics, making it difficult to balance compression rate and semantic preservation in complex scenarios. To address this, we propose a behavior-informed adaptive framework for hierarchical trajectory compression. The framework integrates stay region identification, behavior-oriented segmentation, and multi-feature adaptive compression, enabling differentiated compression across various navigation phases. Stay regions are identified using motion features and spatial density. Navigational behavior patterns are constructed from course sequences, and segmentation is performed using a combination of discrete wavelet transform and entropy-based techniques. Furthermore, introduce multi-dimensional deviation factors and trajectory bending factors, while dynamically setting the compression threshold through a baseline scale adjustment mechanism. In experiments across three representative maritime regions, our method achieves an average compression rate of 78.4 % with a mean spatial error of only 57.8 m, while also maintaining low speed error (0.154 kn) and course error (26.9°). Compared with the benchmark and six advanced algorithms, it consistently delivers the best overall performance, and tests on four typical trajectories further validate its adaptability and robustness.