Change detection in remote sensing images is of great importance as a basis for a variety of tasks. However, traditional fully convolutional change detection networks lack global information, while transformer-based networks can extract global information, but the number of parameters and computational complexity are too large. This paper proposes a lightweight change detection network that fuses global information, with only 2.1M parameters, to address these issues. The network combines the local information extracted by fully convolutional with the global information extracted by LSTM to take full advantage of both. Experiments were carried out on three different types of datasets: LEVIR-CD, SYSU-CD and NJDS. The results show that the IOU of the model was improved by 0.94%, 4.74% and 0.9% respectively. This confirms that the model has a high accuracy with a very small number of parameters and a low computational cost, providing a new solution for efficient change detection.