30 / 2023-03-29 11:20:18

Research on Vision/UWB Fusion Location Model for Indoor Complex Scene

Monocular Vision; UWB; ORB-SLAM; Pedestrian Indoor Location; Scale Ambiguity

Indoor localization and navigation is considered an enabler for a variety of applications, such as guidance of passengers in airports, conference attendees, and visitors in shopping malls, hospitals, or office buildings. Visual localization systems have been promoted as promising positioning solutions for applications in industry, as well as robot and pedestrian localization and navigation. However, there are some challenges for visual localization methods, such as the quality and distinctiveness of the query images, fast motion possibly making the camera-visible scene blurry, and sparse texture, or too bright or too dark optical fiber, which lead to the failure of visual location.

Aiming at the failure of visual SLAM caused by sparse wall textures or changes in light brightness in indoor environments, as well as the scale ambiguity and axis offset problems of the SLAM technology for monocular cameras, we propose a method for determining a pedestrian’s indoor location based on an UWB (ultra-wideband) and vison fusion algorithm. The combination algorithm can take the ranging information of UWB and the results of visual feature extraction and matching as observations. Firstly, an UWB localization algorithm based on EKF (extended kalman filter) is proposed, which can achieve indoor positioning accuracy of 0.3 m.  Secondly, a method to solve scale ambiguity and repositioning of the monocular ORB-SLAM2 (oriented fast and rotated brief-simultaneous localization and mapping) algorithm based on EKF is proposed, which can calculate the ambiguity in real time and can quickly reposition when the vision track fails. Lastly, two experiments were carried out, one in a corridor with sparse texture and the other with the light brightness changing frequently. The results show that the proposed scheme can reliably achieve positioning accuracy on the order of 0.2 m; with the combination of algorithms, the scale ambiguity of monocular ORB-SLAM2 can be solved, with the failed vision trace repositioned by UWB, and the positioning accuracy of UWB can be improved, making it suitable for pedestrian location in indoor environments with sparse texture and frequent light brightness changes.

With the rapid development of urbanization, the demand for emergency rescue in indoor or urban underground space is increasing. Through the combination of vision and UWB, not only can the emergency location of indoor rescue workers be realized, but also the video data of the rescue scene can be obtained in real time, which has great application potential.