In maritime traffic situation, prevention of collision between ships is one of the important factors for the waterway safety. In order to assess ship collision risk, it needs to consider variable factors likes ship traffic quantity, ship type, route i...
In maritime traffic situation, prevention of collision between ships is one of the important factors for the waterway safety. In order to assess ship collision risk, it needs to consider variable factors likes ship traffic quantity, ship type, route information and encounter type and so on. Statistical data of the maritime traffic situation is one of the possible method to assess ship collision risk analysis.
In this study, we derive collision risk assessment and estimation model by means of logit model using ship trajectory.
Firstly, AIS raw data pre-processed for time synchronization and position interpolation. Then it converted to ship encounter data and route ship traffic data.
Secondly, ship encounter data is divided into ship collision variance variables and ship approach variables. For eliminating multicollinearity, factor analysis conducted with these selected variables.
Thirdly, route ship traffic data is used to annual calculate collision candidates by means of the IWRAP MK2 programme. Collision candidates multiplying geometric collision candidates and Causation factor are selected as an another dependent variables.
To do regression analysis, a dependent variable reflecting the objective effect is necessary. In this study, we suggested two dependent variables. One is dichotomous dependent variable which determined Near miss occurrence(Near miss/no Near miss). Dichotomous dependent variable can represent the effect of independent variables by logistic regression. The other is ordered dependent variable which has 5 stepwise Ship domain dimensions level by 4 Ship domains from a ship's position. Ordered dependent variable can derive regression equations by the ordered logit model.
To predict ship collision risk, Near miss level by stepwise Ship domain dimensions are estimated using ordered logit regression equations.
In conclusion, statistical risk assessment based on encounter data using Near miss variable as dependent variables can be utilized in analysis of the ship encounter situation and estimation of the ship Near miss risk. In terms of preparing for the big data era in maritime safety area, this study can be useful for analysis and estimation of the ship collision risk.