Efficient 3D object recognition via geometric information preservation

Abstract

Accurate 3D object recognition and 6-DOF pose estimation have been pervasively applied to a variety of applications, such as unmanned warehouse, cooperative robots, and manufacturing industry. How to extract a robust and representative feature from the point clouds is an inevitable and important issue. In this paper, an unsupervised feature learning network is introduced to extract 3D keypoint features from point clouds directly, rather than transforming point clouds to voxel grids or projected RGB images, which saves computational time while preserving the object geometric information as well. Specifically, the proposed network features in a stacked point feature encoder, which can stack the local discriminative features within its neighborhoods to the original point-wise feature counterparts. The main framework consists of both offline training phase and online testing phase. In the offline training phase, the stacked point feature encoder is trained first and then generate feature database of all keypoints, which are sampled from synthetic point clouds of multiple model views. In the online testing phase, each feature extracted from the unknown testing scene is matched among the database by using the K-D tree voting strategy. Afterwards, the matching results are achieved by using the hypothesis & verification strategy. The proposed method is extensively evaluated on four public datasets and the results show that ours deliver comparable or even superior performances than the state-of-the-arts in terms of F1-score, Average of the 3D distance (ADD) and Recognition rate.

A New Edge Patch with Rotation Invariance for Object Detection and Pose Estimation

“…The average time of our online testing phase on the Tejani dataset was 903.4 ms, which is close to the 774.5 ms in Liu et al [24]. The online testing phase consisted of four stages, namely, 'Patch sampling', 'Feature encoding', 'Hypothesis generation', and 'Hypothesis verification'.…”

“…For the objects with heavy clutter, our average F1-score was 0.95, while those of the methods in [7,16,21,24] were 0.864, 0.872, 0.755, and 0.919, respectively. With the aggravation of background clutter, our average F1-score decreased by 0.012, while those of the methods in [7,16,24] decreased by at least 0.027. These data prove that the proposed E-patch achieved stronger robustness against clutter.…”

“…Considering a synthetic E-patch Psyn and the corresponding realistic E-patch Prel, the relationship between them is expressed as Equation (21): Taking a clutter index of 77.5% as the dividing point, objects were divided into two groups: those with slight clutter ('Joystick', 'Milk', and 'Juice Carton') and those with heavy clutter ('Coffee Cup', 'Shampoo', and 'Camera'). For the objects with slight clutter, the average F1-score of the proposed method was 0.962, while those of the methods in [7,16,21,24] were 0.956, 0.899, 0.74, and 0.959, respectively. The pure white color of 'Milk' made RGB-D patches inside the object too similar to distinguish, leading to failure in [21].…”

“…The F1-scores of the proposed method are compared with those of the state-of-the-art methods in Table 1. The results of the comparison methods were obtained from [7,24]. The proposed method obtained a higher average F1-score (0.956) than did the other methods (0.910, 0.885, 0.747, and 0.939).…”

“…In our online testing phase, the computation times of each stage were 153.9 ms, 16.5 ms, 228.7 ms, and 504.3 ms, respectively, while those in [24] were 12.5 ms, 47.4 ms, 186.2 ms, and 528.4 ms, respectively. In 'Feature encoding' and 'Hypothesis verification', our times were roughly the same as those in [24]. The introduction of a canonical orientation led to longer times in 'Patch sampling' and 'Hypothesis generation', which was acceptable considering the improvement in the detection accuracy.…”