Achieving robust and accurate positioning in an unknown place without a map is an important problem in the field of mobile robotics, and scene recognition is an effective way to solve this problem. In order to improve the mobile robot's ability to recognize indoor scenes at home, inspired by the way that humans look around and observe in the scene, this paper proposes an indoor scene recognition model based on multi-view fusion. With the help of multi-view learning theory, the model realizes the extraction of complementary information and consistent information in different views, which is conducive to making the mobile robot have a comprehensive observation of the scene, and overcome the deficiency of observing from a single angle. This work greatly improves the robustness of scene recognition. In order to verify the performance of the model, we have collected the corresponding multi-view scene datasets. Based on the datasets, we have trained the models and compared it with the current state-of-the-art scene recognition models. The experimental results verified the effectiveness of the model. The accuracy is greatly improved compared to the traditional single-view scene recognition network. At the same time, the model in this paper also shows good robustness to the visual dislocation of the scene.