A Robust Indoor Localization Method Based on DAT-SLAM and Template Matching Visual Odometry

Robust positioning is a central issue in robots. Due to the complex indoor environment, visual simultaneous localization and mapping (VSLAM) is susceptible to light and some scenes contain few features. The robot cannot obtain an accurate position. To solve these problems, this work first proposes dynamic adaptive threshold simultaneous localization and mapping (DAT-SLAM), a method for extracting feature points based on a dynamic adaptive threshold. The method is stable for feature extraction under illumination transformation. Subsequently, template matching visual odometry (VO) is introduced and combined with DAT-SLAM to form a joint indoor localization framework. Experiments on the datasets show that DAT-SLAM has a better performance than oriented fast and rotated brief SLAM2 (ORB-SLAM2). The mean positioning accuracy is improved by 11.74%. Experiments in real scenes show that the joint localization framework can achieve continuous localization in scenes with sparse features.