Wireless sensor network (WSN), as a new information collection method, has the advantages of high data sensing accuracy, good environmental adaptability, large coverage area, good concealment and strong invulnerability. Due to the limitation of sensor nodes' own size and WSN's working environment, the energy of sensor nodes in the network is generally limited, and it can't be replenished after the energy is exhausted, which makes it impossible to deploy and apply WSN on a large scale and for a long time in practice. In this paper, a data acquisition and reconstruction algorithm for the sensor layer of the Internet of Things (IOT) based on back propagation neural network (BPNN) and compressed sensing theory is proposed. Under the premise of ensuring the accuracy of data collection, the redundant information of the original data is removed to the greatest extent by compressing the member nodes and cluster head nodes in the cluster, and the data transmission in the network is reduced. The simulation results show that compared with the WSN data acquisition method in literature [9], the energy consumption of the WSN data acquisition method in this paper is reduced by more than 10%. The proposed algorithm can reduce the reconstruction complexity and reconstruction delay of the fusion center without increasing the compressed observation dimension, thus achieving the purpose of reducing network transmission energy consumption.