Abstract: Time-lapse seismology is one of the most important techniques for reservoir monitoring and subsequent development plan adjustment.In an offshore time-lapse seismic survey, we use a legacy streamer dataset to reconstruct the gaps, which were related to the restriction of platform and subsea pipelines, in the ocean bottom node (OBN) seismic data acquired recently.Prestack consistency processing and poststack global matching are employed to obtain the time shifts which reflect gas features effectively.The nonlinear mapping relationship between two datasets after consistency processing, which is generated using a neural network, could then be used to reconstruct missing OBN data.The method proposed in this paper can recover missing seismic data effectively and meanwhile preserve time-lapse seismic characteristics as much as possible.A field application demonstrates a good result.