Tracking of oil spill extended targets based on random point pattern and GLMB

Abstract

In the marine environment, oil spill surfaces exhibit complex drifting and spreading behavior, which makes it difficult to obtain the number of oil spill sources, as well as important parameters such as the location, shape, drifting speed and expansion rate of oil spill surfaces. Moreover, due to external factors such as sea winds, tides, and currents, multiple oil slicks may overlap and merge, making the accurate tracking of multiple oil spill surfaces even more difficult. To address the aforementioned issues, the proposed an oil spill extended target tracking method based on random point pattern and the Generalized Label Multi-Bernoulli (GLMB) filter. Firstly, to address the challenge of solving the parameters for the mixture distribution, the Expectation-Maximum (EM) algorithm of point pattern is introduced to obtain two types of parameters: cardinality distribution and feature distribution; Secondly, the GLMB filtering method is utilized to track the centroid drift trajectory of the oil spill while simultaneously recording information about the shape and concentration of the oil spill surface. Theoretical and simulation experiments demonstrate that the algorithm is capable of accurately estimating the parameters and state of the oil spill, with a tracking error precision of 0.001°(equivalent to 3.6′′ of Earth’s longitude and latitude). The final tests in the real water environment in the laboratory show that the tracking error of the oil spill surface is 0.4 cm, which proves the effectiveness and practicality of the algorithm.