Abstract: In Orthogonal Time-Frequency Space (OTFS) systems for high-speed mobile communication scenarios, accurate acquisition of channel state information (CSI) is crucial. In this paper, the pilot optimization problem in compressive sensing-based channel estimation is investigated by introducing the minimization of average block coherence (ABC) criterion as the optimization metric, and extending the pilot symbol constellation from conventional BPSK to 16QAM. An adaptive Lévy search with local block refinement pilot optimization algorithm (ALS-BR) is proposed. It is demonstrated by simulation results that, compared with existing algorithms based on the average inner product (AIP) criterion, a 2-4 dB improvement in channel estimation accuracy is achieved by the proposed ALS-BR algorithm, while the system's bit error rate (BER) is effectively reduced.