Owing to increasingly complex surface conditions of land seismic acquisition in petroleum exploration, conventional wired acquisition has become environmentally unfeasible, leading to the dramatic increase in exploration cost. We recover and reconstruct the acquired OBN data by the technology of compressed sensing based on curvelet transform, iterative threshold and sampling matrix according to the acquired data. As per model tests, data restoration is satisfactory in the context of complete data on both sides of missing traces and missing traces less than 6% of total traces. The threshold and curvelet scale will be defined depending on measured data. A field OBN data application with 1378 shots yields good results of missing data restoration and reconstruction. The improved event continuity and information content demonstrate the validity of compressed sensing for data restoration and reconstruction.