FAst Detection Algorithm for 3D Keypoints (FADA-3K)

Shogo Arai,Koichi Hashimoto,Nobuaki Fukuchi

doi:10.1109/access.2020.3025534

Shogo Arai, Koichi Hashimoto + Show 1 more

Open Access

https://doi.org/10.1109/access.2020.3025534

Copy DOI

Abstract

3D keypoint is widely used for object recognition and pose estimation with a 3D point cloud since it has robustness against measurement noise and occlusion. Many types of 3D keypoints and the corresponding detection methods of 3D keypoints have been proposed. It is essentially important to select suitable 3D keypoint depending on a target object, scene situation, and 3D measurement system. While there have been a lot of useful methods for detecting 3D keypoints, detecting time is an issue, especially for real-time applications, such as robot vision. The detecting of 3D keypoints tends to entail a trade-off between computation time and the robustness of the 3D keypoint. To solve these problems, we propose a FAst Detection Algorithm for 3D keypoints named FADA-3K. A user can select one of the suitable 3D keypoints and the corresponding detection method which has been already proposed since FADA-3K can be implemented as an add-on of the existing detection methods. Numerical experiments show that FADA-3K can achieve about nine times faster detection than conventional approaches to detecting 3D keypoints.

Highlights

With advances of three-dimensional techniques, such as active vision [1], [2] and stereo vision [3]–[5], technologies for processing 3D point cloud are rapidly developed in purpose of analysis of human motion [6]–[8] and prediction [9], pose estimation [10], [11], object recognition [12], [13], semantic segmentation [14], [15], robotic bin-picking [16], [17], creating map [18], [19], and visual servoing [20]
This paper focuses on object recognition and pose estimation using 3D keypoints
It is assumed that a 3D model of a target object is given, and a purpose is to recognize and estimate the pose of the target object in a scene

Summary

INTRODUCTION

With advances of three-dimensional techniques, such as active vision [1], [2] and stereo vision [3]–[5], technologies for processing 3D point cloud are rapidly developed in purpose of analysis of human motion [6]–[8] and prediction [9], pose estimation [10], [11], object recognition [12], [13], semantic segmentation [14], [15], robotic bin-picking [16], [17], creating map [18], [19], and visual servoing [20]. Computation cost for the detection of 3D keypoints is relatively high in the recognition and estimation processes Performances of these methods depend on the shape of the target object, scene situation, and 3D measurement system. The 3D keypoints in the model and the corresponding 3D keypoints in the scene are required to be detected as the same points in the object’s coordinate system It means that the detection methods for 3D keypoints have to be robust against measurement noise and changes of relative pose between the 3D sensor and objects. OVERVIEW OF FADA-3K Conventional keypoint detection methods require relatively long computational time because they give judgment on whether each 3D point is a 3D keypoint or not. The first part creates a look-up table that is computed based on the given point cloud of the model in an offline fashion.

FADA-3K

COMPRESS DIMENSION OF 3D FEATURE

Findings

VIII. CONCLUSION