A problem of detection of outside historical routes for radar maritime targets is considered. A new method is proposed for finding of normal historical routes and for detection of abnormal routes using radar data. This method is based on a grid-based modelling, an algorithm for finding of routing angles of all trajectories in each cell, and a modified DBSCAN algorithm with a new introduced distance function of two routing angles. It is shown that the classical distance functions such as Euclidean and Hausdorff cannot be able to use with the DBSCAN algorithm for clustering of routing angles. A test with real data is given. The false alarm rate (FAR) of the test is 0.