Abstract

Due to the lack of GNSS signals in indoor environments, landmarks are particularly important in orientation recognition and navigation. However, the salience evaluation models established by existing studies for extracting landmarks have deficiencies in dealing with indicator redundancy and nonlinear relationships. Therefore, this study adopts a data-driven approach. First, the indoor POI salience evaluation indicators system is established. Then, large shopping malls are selected to collect data, and the indicators are screened by Spearman's correlation coefficient after preprocessing. Finally, the importance ranking of evaluation indicators is obtained, and based on this ranking, the model is trained and optimised using random forest and multilayer perceptron. The results show that the model established in this study has higher salience prediction accuracy and the best stability compared with the models established in previous studies.