Classifying optical remote sensing images is a crucial subtask in remote sensing image interpretation. With the advancement of remote sensing technology, more complex scene data can be acquired. However, remote sensing image classification faces challenges due to inter-class image similarities and intra-class diversities, as well as scene significant scale differences in these scenes. To address these issues, this paper proposes a novel method for remote sensing scene classification using a Multi-scale Local Attention Network (MSLANet). The method integrates dual attention mechanisms of channel and space in feature extraction to enhance model sensitivity to regions of interest. Multi-scale features are also fused, this is an improvement that aimed at better fusing image features to enhance model robustness. Our model achieves state-of-the-art performance on three publicly available scene classification datasets.