We propose an adaptive seamless vertical handover scheme for hybrid visible light communication-infrared (VLC-IR) networks in ship cabins based on Sarsa-lambda reinforcement learning algorithm. Compared with the Q-learning and Sarsa, the proposed algorithm surpasses the average downlink data rate by 13% and 14%, respectively. Moreover, the proposed algorithm outperforms the average downlink data rate of immediate handover and dwell handover schemes by 57% and 22%, respectively when the user device movement speed is 1 m/s.