现代物流已经进入了智慧物流的时代，智能运输作为智慧物流的核心组成部分，涌现了以自动引导车（Automated Guided Vehicle，AGV）为代表的系列智能化运输设备。而AGV叉车的货物托盘检测作为在智能运输过程中的主要问题，要求能够在复杂场景中检测出货物托盘的位置和姿态，这是AGV叉车实现智能运输的关键环节。现有货物托盘检测方法主要依据传感设备的不同分为两类，一类是基于激光雷达的检测方法，另一类是基于视觉传感的检测方法。由于激光雷达价格昂贵且数据采集能力有限，难以满足AGV叉车长时间稳定工作的需求；视觉传感器价格相对低廉且采集能力较强，在智能运输中具有更高的应用价值。

本文在分析现有AGV叉车的货物托盘检测算法基础上，分别提出了基于二维图像和三维点云模型的货物托盘检测算法，有效提高了货物托盘检测的精度与稳定性，主要研究内容包括：

（1）基于统一样本加权预测网络的货物托盘检测算法研究。针对面向AGV叉车的二维货物托盘图像检测存在样本均衡性差的问题，基于样本动态加权的方法，提出一种统一样本加权预测网络模型，利用样本在分类损失、回归损失和概率得分上的分布特性来预测样本权重；并将其与目标检测模型相融合，以统一样本加权预测网络模型的权重预测结果优化目标检测模型的权重，进而设计一种基于统一样本加权预测网络的货物托盘图像检测算法，以提升目标检测算法处理非均衡化样本的能力，实现面向AGV叉车的二维货物托盘图像检测。

（2）融合尺度不变特征变换（Scale-Invariant Feature Transform，SIFT）及三维点云的货物托盘检测方法研究。针对面向AGV叉车的三维货物托盘点云模型检测存在区域特征提取能力弱的问题，结合SIFT特征的方向编码特性和尺度感知特性分别构建面向三维货物托盘点云模型的方向编码单元与尺度感知单元，并通过二者的结合，提取货物托盘的点云方向特征与尺度感知特征；在此基础上，将方向编码单元和尺度感知单元与现有点云模型检测网络相结合，提出一种融合SIFT特征及三维点云的货物托盘检测方法，提升货物托盘点云模型检测网络的特征提取能力，实现面向AGV叉车的三维货物托盘点云模型检测。

（3）面向AGV叉车的货物托盘检测系统设计。结合AGV叉车的实际工作场景，设计并搭建面向AGV叉车的货物托盘检测系统。系统主要包括图像采集模块、二维货物托盘图像检测模块和三维货物托盘点云模型检测模块。通过系统测试实验验证货物托盘检测算法的有效性以及货物托盘检测系统的可行性。

[1] 卢阳光, 闵庆飞, 刘锋. 中国智能制造研究现状的可视化分类综述——基于CNKI(2005-2018)的科学计量分析[J].工业工程与管理, 2019, 24(04): 14-22.

[2] Khatib E J, Barco R. Optimization of 5G Networks for Smart Logistics[J]. Energies, 2021, 14(6): 1758.

[3] Wang Z T, University B W. "Smart Logistics" Is Needed by Urbanization[J]. China Business and Market, 2014, 3(10): 4-8.

[4] Zhou Yongning, Xie Ziyang, Zhang Hao, et al. The Advance and Simulation of AGV Wireless Charging Circuit[J]. Video Engineering, 2021,45(1): 67-72.

[5] 凌忠奇. AGV小车路径规划算法的探究[D]. 北京: 机械科学研究总院, 2013.

[6] 刘海. 智能仓储物流导航AGV叉车. 中国, CN208166507U[P]. 2018.11.30.

[7] 王宇斐. 高速高精度AGV小车的智能动态调度系统研究[D]. 杭州: 浙江大学, 2019.

[8] 王鼎. AGV调度系统与路径规划研究[D]. 济南: 山东大学, 2018.

[9] Cui G Z, Lu L S, He Z D, et al. A robust autonomous mobile forklift pallet recognition[C]// CAR 2010 - 2010 2nd International Asia Conference on Informatics in Control, Automation and Robotics, 2010: 286-290.

[10] Casado F, Lapido Y L, Losada D P, et al. Pose Estimation and Object Tracking Using 2D Images[J].27th International Conference on Flexible Automation and Intelligent Manufacturing, 2017, 11(2351): 63-71.

[11] Fooladivanda A, Chehrerazi N, Sadri S, et al. Automatic segmentation of pallet images using the 2D wavelet transform and YUV color space[C]// Iranian Conference on Electrical Engineering, 2010: 48-52.

[12] Mohamed I S, Capitanelli A, Mastrogiovanni F, et al. Detection，localisation and tracking of pallets using machine learning techniques and 2D range data[J]. Neural Computing and Applications, 2019, 32(13): 8811-8828.

[13] Wang, Ye A, Hao G, et al. Autonomous pallet localization and picking for industrialforklifts based on the line structured light[C]// IEEE International Conference on Mechatronics and Automation, 2016: 252-256.

[14] Zaccaria M, Monica R, Aleotti J. A Comparison of Deep Learning Models for Pallet Detection in Industrial Warehouses[C]// 2020 IEEE 16th International Conference on Intelligent Computer Communication and Processing, 2020: 417-422.

[15] Liu J, Tianjian L I. Research of warehouse pallet detection algorithm based on deep learning[J]. Journal of Beijing Information ence and Technology University, 2017, 32(2): 78-84.

[16] Girshik，Gong, Zhang S. Comprehensive molecular characterization of clear cell renal cell carcinoma with caval tumour thrombus[J]. 34th Annual EAU Congress, 2019, 18(1): 2100.

[17] Liu W, Peng H, Zheng Z, et al. FBCRI Based Real-time Path Planning for Unmanned Aerial Vehicles in Unknown Environments with Uncertainty[J]. Robot, 2013, 35(6): 641-650.

[18] Redmon J, Divvala S, Girshick R, et al. You Only Look Once: Unified, Real-Time Object Detection[C]// IEEE Conference on Computer Vision and Pattern Recognition, 2016: 779-788.

[19] Kai Zhao, Qi Han, Jun Xu, et al. Deep Hough Transform for Semantic Line Detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 10(1109).

[20] Yang Xue, Wang J, Hou Y. A meta‐analysis of the effectiveness of Yangxue Qingnao granules for the treatment of chronic cerebral circulation insufficiency[J]. Brain and Behavior, 2020, 10(5).

[21] 朱睿, 江顺亮, 叶发茂, 等. 基于ImageNet预训练卷积神经网络的遥感图像检索[J]. 武汉大学学报(信息科学版), 2018, 43(001): 67-73.

[22] Manhardt F, Kehl W, Gaidon A. ROI-10D: Monocular Lifting of 2D Detection to 6D Pose and Metric Shape[C]// IEEE Conference on Computer Vision and Pattern Recognition, 2019: 2064-2073.

[23] Wu Z, Larry S. Davis, Sigal L. Weakly-Supervised Spatial Context Networks[C]// 2019 IEEE Winter Conference on Applications of Computer Vision, 2019: 1253-1261.

[24] Yan Y, Duffner S, Phutane P, et al. 2D Wasserstein Loss for Robust Facial Landmark Detection[J]. Pattern Recognition, 2021, 116(10): 104975.

[25] Varga R, Nedevschi S. Robust Pallet Detection for Automated Logistics Operations[C]// Proceedings of the 11th Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, 2016: 470-477.

[26] Wauters T, Christiaens J, Berghe G V. Mixed case 3D pallet loading with optimized robotic loadings[C]// ESICUP, 2013: 1-8.

[27] Arpenti P, Caccavale R, Paduano G, et al. RGB-D Recognition and Localization of Cases for Robotic Depalletizing in Supermarkets[J]. IEEE Robotics and Automation Letters, 2020, 50(99): 6233-6238.

[28] Monica R, Aleotti J, Rizzini D L. Detection of Parcel Boxes for Pallet Unloading Using a 3D Time-of-Flight Industrial Sensor[C]// IEEE International Conference on Robotic Computing, 2020: 314-318.

[29] Molter B, Fottner J. Real-time Pallet Localization with 3D Camera Technology for Forklifts in Logistic Environments[C]//Proceeding of the 2018 IEEE International Conference on Service Operations and Logistics, and Informatics. 2018: 297-302.?

[30] Wu D, Cao W, Zhu Y, et al. Research on Pallet Detection Technology Based on 3D Point Cloud and Image Edge Features[J]. Automation & Information Engineering, 2019, 78(55): 93-98.

[31] Liu W, Gao Y, Wang Y, et al. A Novel Edge Detection and Localization Method of Depalletizing Robot[C]// International Conference on Intelligent Robotics and Applications, 2020: 512-521.

[32] Wenhan W U, Yang M, Wang B, et al. Pallet Detection Based on Contour Matching for Warehouse Robots[J]. Journal of Shanghai Jiaotong University, 2019, 58(90): 88-95.

[33] Zhang T, Zhong Y, Song J. RGB-D Saliency Detection with Multi-feature-fused Optimization[C]//International Conference on Image and Graphics, 2017: 15-26.

[34] Hu J F, Zheng W S, Lai J H, et al. Jointly Learning Heterogeneous Features for RGB-D Activity Recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017:39(11): 2186-2200.

[35] Riegler G, Ulusoy AO, Geiger A. OctNet: Learning Deep 3D Representations at High Resolutions[C]// IEEE Conference on Computer Vision and Pattern Recognition, 2017: 6620-6629.

[36] Gupta S, Girshick R, P Arbeláez, et al. Learning Rich Features from RGB-D Images for Object Detection and Segmentation[C]// European Conference on Computer Vision, 2014: 345-360.

[37] Sun S, Zhang T, Li Q, et al. Fault Diagnosis of Conventional Circuit Breaker Contact System Based on Time-Frequency Analysis and Improved AlexNet[J]. IEEE Transactions on Instrumentation and Measurement, 2021(70): 1-12.

[38] Luo J, Jiang M, Liu X, et al. RGB-D object recognition based on multimodal deep learning[J]. Computer Engineering and Design, 2017, 80(88): 4311-4317.

[39] Song S, Xiao J. Deep Sliding Shapes for Amodal 3D Object Detection in RGB-D Images[C]// IEEE Conference on Computer Vision and Pattern Recognition, 2016: 808-816.

[40] 王宁波.基于RGB-D的行人检测[D]. 杭州: 浙江大学, 2013.

[41] 王剑楠.基于RGB-D图像的SLAM问题关键技术研究[D]. 南京: 南京航空航天大学, 2017.

[42] Shotton J, Glocker B, Zach C, et al. Scene Coordinate Regression Forests for Camera Relocalization in RGB-D Images[C]// IEEE Conference on Computer Vision and Pattern Recognition, 2013: 2930-2937.

[43] Pham T, Do T T, N Sünderhauf, et al. SceneCut: Joint Geometric and Object Segmentation for Indoor Scenes[C]// 2018 IEEE International Conference on Robotics and Automation, 2018: 3213-3220.

[44] Chen X, Ma H, Wan J, et al. Multi-View 3D Object Detection Network for Autonomous Driving[C]// IEEE Conference on Computer Vision and Pattern Recognition, 2017: 6526-6534.

[45] Zhou Y, Luo D, Wu B, et al. The active vibration isolation system basedon LADRC algorithm for atom interferometry[J]. Applied Optics, 2020, 59(11): 3487-3493.

[46] Garcia-Garcia A, F Gomez-Donoso, Garcia-Rodriguez J, et al. PointNet: A 3D Convolutional Neural Network for real-time object class recognition[C]// Proceedings of the International Joint Conference on Neural Networks, 2016: 1578-1584.

[47] 王鑫,孙韶媛,顾立鹏, 等. 基于改进Frustum PointNet的3D目标检测[J]. 激光与光电子学进展, 2020, 57(20): 328-334.

[48] Li J, Hu Y. A Density-Aware PointRCNN for 3D Objection Detection in Point Clouds[J]. arXiv preprint arXiv: 2009.05307, 2020.

[49] Li W, Song P. A modified ICP algorithm based on dynamic adjustment factor for registration of point cloud and CAD model[J]. Pattern Recognition Letters, 2015, 65(167): 88-94.

[50] Shuai H, Xu X, Liu Q. Backward Attentive Fusing Network with Local Aggregation Classifier for 3D Point Cloud Semantic Segmentation[J]. IEEE Transactions on Image Processing, 2021, 30(1109): 4973-4984.

[51] H Amano, Ichihashi Y, Kakue T, et al. Reconstruction of a Three-dimensional Color-video of a Point-cloud Object using the Projection-type Holographic Display with a Holographic Optical Element[J]. Optics Express, 2020, 28(4): 5692-5705.

[52] Qi C R, Litany O, He K, et al. Deep Hough Voting for 3D Object Detection in Point Clouds[C]// International Conference on Computer Vision, 2019: 9276-9285.

[53] Engelcke M, Rao D, Wang D Z, et al. Vote3Deep: Fast Object Detection in 3D Point Cloud Using Efficient Convolutional Neural Networks[J]. arXiv preprint arXiv: 1609.06666, 2016.

[54] Hu J F, Zheng W S, Lai J H, et al. Jointly Learning Heterogeneous Features for RGB-D Activity Recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(11): 2186-2200.

[55] Kashefi A, Rempe D, Guibas L J. A point-cloud deep learning framework for prediction of fluid flow fields on irregular geometries[J]. Physics of Fluids, 2021, 33(2).

[56] Chen S, Liu B, Feng C, et al. 3D Point Cloud Processing and Learning for Autonomous Driving: Impacting Map Creation, Localization, and Perception[J]. IEEE Signal Processing Magazine, 2021, 38(1): 68-86.

[57] Zheng K, Yao Y. Automatic three-dimensional reconstruction of subsurface defects by segmenting ultrasonic point cloud[J]. Journal of the Taiwan Institute of Chemical Engineers, 2021, 120(10): 24-32.

[58] Duckett T, Yan Z, Bellotto N. Online learning for 3D LiDAR-based human detection: experimental analysis of point cloud clustering and classification methods[J]. Autonomous Robots, 2020, 44(2): 147-164.

[59] Zhao Y, Kwon K C, Erdenebat M U, et al. Implementation of full-color holographic system using non-uniformly sampled 2D images and compressed point cloud gridding[J]. Optics Express, 2019, 27(21): 29746-29758.

[60] Martinez J, Albeaino G, Gheisari M, et al. UAS Point Cloud Accuracy Assessment Using Structure from Motion-Based Photogrammetry and PPK Georeferencing Technique for Building Surveying Applications[J]. Journal of Computing in Civil Engineering, 2021, 35(1):1-15.

[61] Lza B, Jfa C, Long H A. Semantic segmentation of sparse 3D point cloud based on geometrical features for trellis-structured apple orchard - ScienceDirect[J]. Biosystems Engineering, 2020, 196(68): 46-55.

[62] Tachella J, Altmann Y, Mellado N, et al. Real-time 3D reconstruction from single-photon lidar data using plug-and-play point cloud denoisers[J]. arXiv preprint arXiv : 1905.06700, 2019.

[63] Qi C R, Li Y, Hao S, et al. PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space[C]// IEEE Conference on Computer Vision and Pattern Recognition, 2017: 5100-5109.

[64] Wang W, Yu R, Huang Q, et al. SGPN: Similarity Group Proposal Network for 3D Point Cloud Instance Segmentation[C]// IEEE Conference on Computer Vision and Pattern Recognition, 2018: 2569-2578.

[65] Ren S, He K, Girshick R, et al. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149.

[66] Liu W, Anguelov D, Erhan D, et al. SSD: Single Shot MultiBox Detector[C]// European Conference on Computer Vision, 2016: 21-37.

[67] Bochkovskiy A, Wang C Y, Liao H. YOLOv4: Optimal Speed and Accuracy of Object Detection[J]// arXiv preprint arXiv: 2004.10934, 2020.