关键词: BDSBAS, 智能手机, YOLOv5, 单目测距, 坐标正算

Abstract: To address the insufficient performance analysis of smartphone image target localization,this paper proposes a smartphone image target localization method based on BDSBAS and monocular ranging.BDSBAS-enhanced data is used to improve the positional accuracy of the smartphone as the starting point.YOLOv5 target detection and the principle of similar triangles are used to measure the distance to the target.Yaw angle information from the phone's orientation sensor is fused to determine the target's azimuth angle,thus constructing a forward model for image target coordinate calculation.Experiments show that after BDSBAS enhancement,the horizontal positioning accuracy of a OnePlus 13/vivo X80 phone is within 5 m,and the static and dynamic horizontal positioning accuracy is improved by 16.09% and 36.97% respectively compared to single-point positioning.The monocular ranging accuracy for image targets is within 3 m,and the close-range accuracy can reach the centimeter level.The proposed method can achieve non-contact image target localization relying solely on the smartphone's own sensors.The target planar positioning accuracy of approximately 6 m is comparable to the single-point positioning accuracy of the smartphone.

Key words: BDSBAS, smartphone, YOLOv5, monocular ranging, coordinate calculation