In order to improve the wireless sensor network(WSN) data collection accuracy, reduce the energy consumption of the network and improve the robustness of data collection algorithm under packet loss condition, a data collection scheme based on expected network coverage and cluster compressive sensing is proposed. The data collection scheme is divided into two steps as expected network coverage optimization and cluster CS (compressive sensing) data collection. Firstly, the expected network coverage optimization algorithm is designed, and the node scheduling strategy is given through the quantitative analysis of the node coverage redundancy and the expected value of network coverage in the key observation area, which helps to achieve the purpose of the ``special'' area observation and reduce energy consumption. Then, by analyzing the relationship between networks clustering and node deployment, the adaptive dynamic network clustering results are provided. On this basis, the weak correlation observation matrix is designed, which can reduce the influence of the packet loss on CS data collection. Finally, the social spider optimization algorithm is introduced to improve the reconstruction accuracy of the CS. The simulation results show that compared with other data collection algorithms, the data reconstruction error is reduced by about 23.5{%