Abstract
In recent years, the pose estimation of objects has become a research hotspot. This technique can effectively estimate the pose changes of objects in space and is widely used in many mobile devices, such as AR/VR. At present, mainstream technologies can achieve high‐precision pose estimation, but the problem of that of multiple irregular objects in mobile and embedded devices under limited resource conditions is still challenging. In this paper, we propose a FastQR algorithm that can estimate the pose of multiple irregular objects on Renesas by utilizing homography method to solve the transformation matrix of a single QR code and then establish the spatial constraint relationship between multiple QR codes to estimate the posture of irregular objects. Our algorithm obtained a competitive result in simulation and verification on the RZ/A2M development board of Renesas. Moreover, the verification results show that our method can estimate the spatial pose of the multiobject accurately and robustly in distributed embedded devices. The average frame rate calculated on the RZ/A2M can reach 28 fps, which is at least 37 times faster than that of other pose estimation methods.
Highlights
With the development of AR/VR and control technology, object pose estimation is needed in manifold scenes
A large number of low-cost small embedded platforms need to carry out object spatial pose estimation, such as in the exhibition, education, and other fields [1, 2]
This paper proposes a FastQR algorithm that can estimate the pose of multiple irregular objects on Renesas by utilizing QR codes and a monocular camera and estimate the posture of irregular objects
Summary
With the development of AR/VR and control technology, object pose estimation is needed in manifold scenes. Most of the estimation of object pose depends on the binocular camera [3] or RGB-D camera [4]. This type of camera is large and requires a computing platform with a certain amount of computing power, which is not conducive to the installation on small devices. Using convolution neural network to estimate the position and pose of objects [7] has become a research focus, which brings accuracy, but requires high computational power of the platform, and even needs to be accelerated by GPU
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
