Seismic data acquisition serves as a critical foundation for seismic exploration. Since it accounts for the majority of the costs in seismic exploration, the optimization of acquisition technology is of great significance to the entire seismic exploration workflow. Constrained by the Nyquist sampling theorem, traditional acquisition methods require high sampling rates, lack flexibility in adapting to actual field conditions, and incur substantial exploration costs. Compressed sensing theory has been introduced into seismic exploration for more than a decade. Acquisition methods based on this theory allow for random deployment of receivers, reduce operational time, lower costs, and improve the efficiency of seismic exploration. However, existing compressed sensing based acquisition design methods cannot guarantee a consistent number of traces per shot, which adversely affects reconstruction quality. This paper proposes a two-step sampling method to overcome the inconsistency in the number of sampled traces per shot. The first step employs regular sampling to generate fixed sampling positions and control the spacing between adjacent receivers. The second step introduces random sampling to fill the gaps within the fixed scheme, enhancing the randomness of sampling, and subsequently generates sampling schemes that maintain a consistent number of traces per shot. Data experiments demonstrate that the reconstruction results from the two-step acquisition method achieve a higher signal-to-noise ratio compared to those from random sampling and piece-wise random sampling on average. Therefore, this method holds practical application value in seismic exploration.