Container lock pins recognition based on a self-built dataset with high completeness

doi:10.3969/j.issn.1000-5641.2025.04.003

Abstract

Abstract:

The recognition accuracy of container lock pins is a main factor affecting the efficiency of automatic dismantling of terminals. However, the shape and structure of container lock pins are different, and traditional methods, such as hand-made descriptors for lock pin recognition, are slow and inaccurate. This study proposes a 3D point cloud recognition method for container locking pins based on self-built datasets with high completeness; additionally, a 3D point cloud dataset of container locking pins with high completeness is constructed by realizing a semi-automatic collection system for 3D point cloud data collection of many container locking pins. Based on the self-built dataset, a point cloud classification deep learning network algorithm is used to train the model to verify the rationality of the method of constructing the dataset and the effectiveness of the dataset. The results indicate that the lock-pin recognition method based on the self-built dataset with high completeness has high accuracy. Moreover, the recognition accuracy of a complete single-view 3D point cloud can reach 100%, it is not easily affected by external environmental factors, and it has good performance in the case of missing parts of the point cloud. The research results can provide theoretical and technical support for the automatic identification and disassembly of container lock pins in terminals.

Key words: container lock pins, self-managed dataset, 3D point cloud, lock identification

CLC Number:

TP391

Xingxing XU, Chang HUANG. Container lock pins recognition based on a self-built dataset with high completeness[J]. J* E* C* N* U* N* S*, 2025, 2025(4): 28-37.

Figures/Tables 11

Table 1

Fig.1

Fig.2

Table 2

Fig.3

Table 3

Table 4

Fig.4

Fig.5

Table 5

Table 6

References 20

1	陆拥军.. 码头集装箱锁销自动化拆装技术研究. 港口装卸, 2019, (2): 16- 19.
2	何继红.. 自动化集装箱码头装卸工艺系统应用现状与展望. 水运工程, 2018, (6): 199- 203.
3	邹鲁, 赵永新, 王西超, 等.. 基于深度卷积神经网络的集装箱锁销识别研究. 上海电机学院学报, 2019, 22 (4): 193- 197.
4	汪兆冉, 李保江, 王西超, 等.. 基于深度学习的集装箱锁销识别系统. 机械设计与研究, 2022, 38 (1): 186- 190.
5	ZHOU R, SUN H, MA K S, et al.. Improving estimation of tree parameters by fusing ALS and TLS point cloud data based on canopy gap shape feature points. Drones, 2023, 7 (8): 524.
6	TAMATA K, MASHITA T.. Feature description with feature point registration error using local and global point cloud encoders. IEICE Transactions on Information and Systems, 2022, 105 (1): 134- 140.
7	LIANG Z M, HUANG Y P.. Survey on deep learning-based 3D object detection in autonomous driving. Transactions of the Institute of Measurement and Control, 2023, 45 (4): 761- 776.
8	VANIAN V, ZAMANAKOS G, PRATIKAKIS I.. Improving performance of deep learning models for 3D point cloud semantic segmentation via attention mechanisms. Computers & Graphics, 2022, 106, 277- 287.
9	ZHUANG C G, LI S F, DING H.. Instance segmentation based 6D pose estimation of industrial objects using point clouds for robotic Bin-picking. Robotics and Computer-Integrated Manufacturing, 2023, 82, 102541.
10	HAN X F, SUN S J, SONG X Y, et al. 3D point cloud descriptors in hand-crafted and deep learning age: State-of-the-art [EB/OL]. (2020-07-27)[2023-12-11]. https://arxiv.org/pdf/1802.02297.pdf.
11	TOMBARI F, SALTI S, DI STEFANO L. Unique signatures of histograms for local surface description [C]// Computer Vision–ECCV 2010. 2010: 356-369.
12	PRAKHYA S M, LIU B B, LIN W S. B-SHOT: A binary feature descriptor for fast and efficient keypoint matching on 3D point clouds [C]// 2015 IEEE/RSJ International Conference on Intelligent Robots and Systems. 2015: 1929-1934.
13	RUSU R B, BLODOW N, MARTON Z C, et al. Aligning point cloud views using persistent feature histograms [C]// 2008 IEEE/RSJ International Conference on Intelligent Robots and Systems. 2008: 3384-3391.
14	CHARLES R Q, HAO S, MO K C, et al. PointNet: Deep learning on point sets for 3D classification and segmentation [C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2017: 77-85.
15	QI C R, YI L, SU H, et al. PointNet++: Deep hierarchical feature learning on point sets in a metric space [C]// Proceedings of the 31st International Conference on Neural Information Processing Systems. 2017: 5105-5114.
16	HU H, YU J C, YIN L, et al. An improved PointNet++ point cloud segmentation model applied to automatic measurement method of pig body size [J]. Computers and Electronics in Agriculture, 2023, 205: 107560.
17	WANG W Y, ZHOU H F, YAN Y X, et al.. An automatic extraction method on medical feature points based on PointNet++ for robot-assisted knee arthroplasty. The International Journal of Medical Robotics and Computer Assisted Surgery, 2023, 19 (1): e2464.
18	LIU B J, CHEN S X, HUANG H G, et al.. Tree species classification of backpack laser scanning data using the PointNet++point cloud deep learning method. Remote Sens, 2022, 14 (15): 3809.
19	YAN X, ZHENG C D, LI Z D, et al. PointASNL: Robust point clouds processing using nonlocal neural networks with adaptive sampling [C]// Proceedings of the 2020 IEEE Conference on Computer Vision and Pattern Recognition. 2020: 5589-5598.
20	CHEN G Y, WANG M L, YANG Y, et al. PointGPT: Auto-regressively generative pre-training from point clouds [EB/OL]. (2023-05-23)[2023-12-12]. https://arxiv.org/pdf/2305.11487.pdf.

数据集名称	样本总数/个	类别/种	特点
ModelNet10	4899	10	在 3D CAD(Computer Aided Design‌)模型上采集, 无背景, 无噪声
ModelNet40	12311	40	在 3D CAD(Computer Aided Design‌)模型上采集, 无背景, 无噪声
ShapeNet Core55	51162	55	大规模, 标注丰富

曝光时间/s	增益/dB	触发频率/Hz	速度/(rad·s⁻¹)	与电动转台距离/cm
1000	5	55	30	80

测试集名称	视觉传感器速度/(rad·s⁻¹)	与电动转台距离/cm	与视觉传感器中心位移/cm
测试集 1	30	80	0
测试集 2	60	80	0
测试集 3	30	120	0
测试集 4	60	80	向左20
测试集 5	30	120	向右20

数据集名称	样本量/个	真实性	全面性	样本关联性	样本浓度比例
集装箱锁销点云	21147	通过视觉传感器实际扫描得到	共19个类别, 一个锁销分为一个类别, 每个锁销选取3个姿态, 每个姿态采集371个不同数据	通过转台每1°采集一个数据	每个类别均包含1113个样本
ModelNet40	12311	通过CAD模型采集	共40个大类别, 每个类别由多个CAD模型组成, 每个模型采集十多个数据	CAD模型旋转30°采集一个数据	每个类别约包含300个样本, 不同类别间有较小差异

算法	测试集名称	准确率/%
算法	测试集名称	无预处理OA	经预处理OA
PointNet	测试集 1	99.51	100.00
	测试集 2	46.71	100.00
	测试集 3	12.89	100.00
	测试集 4	46.14	100.00
	测试集 5	12.34	100.00
	ModelNet40	91.40	91.40
PointNet++ -SSG	测试集 1	99.91	100.00
	测试集 2	82.70	100.00
	测试集 3	41.23	100.00
	测试集 4	82.50	100.00
	测试集 5	41.25	100.00
	ModelNet40	92.41	92.41
PointNet++ -MSG	测试集 1	99.95	100.00
	测试集 2	84.37	100.00
	测试集 3	37.01	100.00
	测试集 4	84.22	100.00
	测试集 5	36.46	100.00
	ModelNet40	92.80	92.80
PointASNL	测试集 1	99.94	100.00
	测试集 2	84.78	100.00
	测试集 3	42.48	100.00
	测试集 4	84.73	100.00
	测试集 5	42.63	100.00
	ModelNet40	93.22	93.22
PointGPT	测试集 1	100.00	100.00
	测试集 2	86.94	100.00
	测试集 3	48.68	100.00
	测试集 4	87.05	100.00
	测试集 5	47.94	100.00
	ModelNet40	94.70	94.70