An Indoor-Outdoor Scene Recognition Method Based on AdaBoost and Decision Trees

To address the challenge of accuracy in floor-level localization within multi-building, multi-floor environments, this paper proposes an indoor-outdoor scene recognition method based on the AdaBoost algorithm using the open-source Wi-Fi fingerprint dataset UJIIndoorLoc. By preprocessing and normalizing the Wi-Fi signal data, the method enhances data separability. Additionally, the weight updating mechanism of weak learners and iterative optimization across multiple rounds improve the model's overall classification performance, especially for hard-to-classify samples. Experimental results demonstrate that the proposed method achieves a detection accuracy of 97.05%, outperforming traditional algorithms such as KNN, Bagging, and DNN.