A novel bidirectional selection dynamic time warping algorithm (DTWA) is proposed for similarity analysis of fishing vessel navigation trajectories (FVNT). This algorithm selects data bi-directionally based on the minimum path principle at the beginning and end of the distance matrix, retaining sub-matrices with high utilization rates to reduce computational costs while obtaining the cost of all path selections. Path selection imposes monotonicity, boundary, and continuity constraints to merge the bidirectionally selected paths, and all local shortest paths are superimposed to obtain a more accurate warping path. The bidirectional selection DTW (BSDTW) algorithm calculates FVNT datasets from AIS and BeiDou navigation satellite systems with the KNN algorithm. Results show that our algorithm achieves higher accuracy and lower time cost than traditional DTWA.