Pose visual detection method for cellphone dropping process incorporating prior information

Xuesong Wang,Zheng Zhang,Dongsheng Zhang,Limin Ding

DOI: https://doi.org/10.1007/s11227-023-05527-2

IF: 3.3

2023-08-29

The Journal of Supercomputing

Abstract:The failure mode analysis of the process of dropping a cellphone is significantly impacted by changes in the phone’s pose. During the drop process of a phone, the quality of the IMU sensor pose estimation is challenging to assess due to sensor noise and accelerometer failure induced by weightlessness. Current optical pose recognition methods have limited accuracy and long solution times. In this study, a rapid binocular visual posture solution approach is proposed based on the shape and size of a checkerboard grid and phone as prior information. First, to increase the pose estimation performance, a straightforward and simple checkerboard grid is employed, whose corner points are taken as feature points. Second, to limit the impact of camera distortion and increase pose estimation accuracy, the prior information of the mobile phone and checkerboard grid shape and size is leveraged to correct the corner points’ spatial position information. The experimental findings demonstrate that the recognition outcomes of this detection system are consistent with the actual rotation process, and the relative accuracy is better than ±1∘$$^\circ $$. The proposed method is, therefore, suitable for determining the phone’s pose information in the event of drops with high accuracy and at high speeds, and thus satisfy the criteria for IMU sensor evaluation.

computer science, theory & methods,engineering, electrical & electronic, hardware & architecture