Indoor and Outdoor Scene Classification Method Based on Improved Shared K -Nearest Neighbor

Indoor and Outdoor Scene Classification Method Based on Improved Shared K -Nearest Neighbor

연구 분야: Infrastructure

학회: 2024 20th International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery (ICNC-FSKD)

In view of the problems of insufficient classification accuracy, low reliability and poor stability of current indoor and outdoor scene recognition methods, a method of indoor and outdoor scene classification based on improved shared K-nearest neighbors was proposed. The method makes full use of WiFi Received Signal Strength Indicator (RSSI), number of visible satellites and Signal to Noise Ratio (SNR) The different characteristics displayed in indoor and outdoor environments enable effective scene identification. Firstly, K adjacent points with the highest similarity were selected by calculating the category belonging degree between the points to be classified and the classified samples. Secondly, the sample that is most similar to the point to be classified is identified, and K classified nearest neighbors of the sample are determined. Finally, the improved shared K-neighbor algorithm is used to calculate the number of shared neighbors between the points to be classified and the most similar points, and according to the indoor and outdoor category belonging degree and size of these shared neighbors, the prediction category of the points to be classified is used to achieve the final classification of the classification points. Experimental results show that compared with the traditional classification methods K-Nearest Neighbor (KNN), random forest, decision tree, C4.5 decision tree and Naive Bayesian Model (NBM), the classification accuracy of indoor and outdoor scenes is improved by 1.5%, 9%, 9.5% and 1 % respectively.