【作者】 刘超；

【导师】 张勇； 王国强；

【作者基本信息】 华南理工大学 ， 机械（专业学位）， 2023， 硕士

【摘要】 随着电子商务的快速发展,我国的物流运输量激增,传统物流行业面临着向现代化智慧物流转型的境况。智慧物流体系的建立离不开数据的支持,而物流运输车辆的装载数据是其中的重要一环。传统对于货车车厢装载状态的判断多为人工检测,测量不便且难以给出完整的车厢装载信息。双目视觉的体积测量技术已在部分工业领域得到应用,但多为简单背景下的测量任务。针对车厢内部检测的复杂情景,本文给出一种针对车厢内部装载状态的测量方案,包括货车车厢尺寸的测量和车厢内部装载状态的测量两部分。首先,依据相机成像模型和标定原理对双目相机进行标定实验,使用标定后的双目相机从侧后方采集货车车厢图像,结合现有文献对轮廓提取的分析和货车车厢在图像中的成像特点,选择背景差分、自动Grab Cut和U~2-Net算法作为车厢轮廓提取的候选方法,经实验对比分析后,确定U~2-Net显著性目标检测算法进行车厢轮廓提取,通过改进的基于图像分块的直线检测算法提取车厢边缘,结合车厢轮廓特征,利用车厢边缘直线计算出特定的角点坐标,由三维坐标恢复后的车厢角点间距得到车厢尺寸。对某款厢式货车进行测量实验,验证了提出的车厢尺寸测量方法的可行性。在上述研究的基础上,结合车厢的特征,使用基于货物轮廓表面像素点视差值的货物体积计算方法对车厢内部的货物体积进行测量,通过改进的货物轮廓提取方法,进行货物轮廓表面提取。利用SGBM算法计算货物的视差图,并使用具有良好边缘保持性的WLS滤波算法结合左右一致性检测进行视差优化。对货物轮廓表面的视差值进行统计,根据测量模型计算出车厢内部货物体积和当前车厢的装载率,并构建货物在车厢内部码放情况的三维图,结合装载数据输出综合的车厢装载状态信息。在实验室环境下,对车厢的装载情况进行模拟,展开测量实验并分析结果。最后,根据测量流程编写GUI软件,对实车的货物装载状态进行测量实验,能够得到货车装载的详细信息,具有一定的应用前景。更多还原

【Abstract】 With the rapid development of e-commerce,China’s logistics transportation volume has surged,and the traditional logistics industry is facing the need for transformation to modernized,intelligent logistics.The establishment of an intelligent logistics system relies heavily on the support of data,and the loading data of logistics vehicles is an essential component.Traditional methods for determining the loading status of van often rely on manual inspection,which is inconvenient and unable to provide complete loading information.While binocular vision volume measurement technology has been applied in some industrial fields,it has mostly been for simple tasks.For the complex scenario of carriage interior inspection,we proposes a measurement scheme for carriage interior loading status,which includes two parts:measurement of carriage dimensions and measurement of carriage interior loading status.First,based on the camera imaging model and calibration principle,the binocular camera is calibrated.Then,the calibrated binocular camera is used to capture images of the van from the side and rear.To extract the carriage’s contour,background difference algorithm,automatic Grab Cut algorithm,and U~2-Net algorithm are chosen as candidate methods,in combination with the analysis of contour extraction in existing literature and the imaging characteristics of the carriage in images.The U~2-Net saliency target detection algorithm is used for the extraction of carriage contours,and the carriage edges are obtained by an improved straight line detection algorithm based on image chunking.The specific corner point coordinates are calculated by using the carriage edge straight lines in combination with the carriage contour features,and the carriage dimensions are obtained from the carriage corner point spacing after the 3D coordinate recovery.The feasibility of the proposed carriage dimension measurement method is verified by conducting measurement experiments on a van.Based on the above study,the cargo volume inside the carriage is measured using a cargo volume calculation method based on the parallax value of the pixel points on the cargo contour surface by an improved cargo contour extraction method,combined with the characteristics of the carriage.The parallax map of the cargo is calculated by the SGBM algorithm,and parallax optimization is performed by the WLS filtering algorithm combined with left-right consistency check.The parallax value of the cargo contour surface is counted,and the volume of cargo inside the carriage and the current loading rate of the carriage are calculated according to the measurement model.A three-dimensional map of the cargo inside the carriage is constructed,and comprehensive information on the loading status of the carriage is output in combination with the loading data.In the laboratory environment,we simulate the loading status of the carriage,conduct measurement experiments,and analyze the results.Finally,the GUI software is written according to the measurement process,and the measurement experiment is conducted on the loading status of a real van.This approach can obtain detailed information on the loading status of the van and has certain application prospects.更多还原