| (i) | Input: Samples to be clustered , sailing radius , the smallest number of vessels in a cluster , number of behaviors |
| (ii) | Output: clustering results |
| (1) | Mark all points in as unvisited//Start DBSCAN clustering |
| (2) | Calculate the matrix , with each cell representing the Mercator distance between each two points |
| (3) | Do |
| (4) | Randomly select an unvisited point |
| (5) | Mark as visited |
| (6) | Initialize |
| (7) | If there are at least points in field of , then//The mercator distance between two points can be found in the |
| (8) | Initialize, add to |
| (9) | Let be the points set in the field of |
| (10) | For each in |
| (11) | If is unvisited, then |
| (12) | Mark as visited |
| (13) | If there are at least points in the field of , then |
| (14) | Add points to |
| (15) | End If |
| (16) | If is not a member of any cluster, then |
| (17) | Add to |
| (18) | End If |
| (19) | End If |
| (20) | End For |
| (21) | Add to |
| (22) | Else mark as noise point |
| (23) | End If |
| (24) | Until all the points are marked, //DBSCAN clustering is complete |
| (25) | Compute the average speed set of each in |
| (26) | Select points as the initial center point: //Start KMeans clustering |
| (27) | Do |
| (28) | Initialize |
| (29) | For in do |
| (30) | Compute the speed difference between and |
| (31) | The cluster label of was determined according to the nearest cluster center |
| (32) | Add to the nearest cluster: |
| (33) | For do |
| (34) | Compute the new cluster center: |
| (35) | If, then |
| (36) | Update as the cluster center |
| (37) | End If |
| (38) | End For |
| (39) | End For |
| (40) | Until the update of all clusters is complete//KMeans clustering is complete |
| (41) | Return //Behavior classification is complete |
