Monocular Vision Aided Inertial Positioning Method based on Centroid Extraction Applicable to GNSS Denied Environments

Focusing on the issue that the continuous integrated inertial/visual positioning method requires high hardware requirements, which is adverse to low-cost applications, a monocular vision aided inertial positioning method is proposed. A vision aided inertial positioning error correction model is established, and the spatial relationship between the camera and the marker's centroid is obtained through visual motion estimation. On the premise that the coordinate of the marker's centroid is known, the inertial positioning error can be corrected. Taking the GPS trajectory as the ground truth and the sign on the pavement as the road marker, the field experiment was carried out by using the MEMS IMU and the monocular camera. Our results show that the positioning error decreases from several hundred meters to nearly 2 meters after adding visual information. Experiment demonstrates that the proposed method is valid, which provides a reference for positioning under the GNSS denied environment.