多工位协同零部件位姿感知数据集
收藏资源简介：
数据内容：数据内容主要有翼面吊装零件标准模型数据、翼面鸭翼轴位姿估计Linemod数据集、动态可视化装配引导策略与多人交互模块数据、多人协同的智能决策方法与任务分发机制数据集。用于智能辅助装配的多工位协同机制，实现多人协同的智能决策方法与任务分发机制，通过装配引导策略，构建高效自适应引导装配新模式，实现复杂多源场景下多人交互内容的准确表达与反馈，实现零部件位姿的精确感知，最后通过增强现实可视化装配自适应引导，实现多人协同的高效精准操作。翼面吊装零件标准模型数据包括蒙皮、骨架及鸭翼轴STEP文件，以及鸭翼轴和装配箱体的FBX、PLY文件；翼面鸭翼轴位姿估计Linemod数据集包括深度图(depth)、彩色图(RGB)、标签(labels)、掩码图(mask)、分割图(segnet_results)、位姿变换矩阵(transforms)、模型文件(ply)、图像真值信息(gt.yml)、相机内参向量表示(info.yml)、相机内参及深度比例因子(intrinsics.json)，训练和测试txt文件；动态可视化装配引导策略与多人交互模块数据主要包括源代码前后30页word文档、动态可视化装配引导策略与多人交互模块说明书，鸭翼轴和装配箱体FBX模型文件；多人协同的智能决策方法与任务分发机制数据集主要包括obj文件、预制件prefab文件以及多人协同的智能决策方法与任务分发机制报告。 采集方案：测量实际模型的关键几何尺寸，在SolidWorks中设计模型的结构，并在装配图中试装全部零件，在零件体中新建坐标系设置XYZ，将设计好的零件模型导出为STEP文件，在SolidWorks中将关键部件鸭翼轴转换为Ascii编码的PLY格式，将文件导入3Dmax将格式转换为FBX格式数据；在场景中围绕目标均匀布置Aruco码定位，平稳移动云台，录制采集目标各方位数据，包括RGB和RGB-D图，利用相机内参将深度图转为点云。基于点云重构计算初始位姿，并对初始点云进行滤波，用高精度模型替换初始点云三角化后的模型，拼接多段数据集，生成目标物体掩码，生成二维与三维包围框，最后得到完整格式的数据集；在Unity中导入为模型和预制件。 体量：大约4.0GB
Data Content: The dataset mainly includes four parts: standard model data of wing hoisting parts, Linemod dataset for wing canard shaft pose estimation, data of dynamic visual assembly guidance strategy and multi-person interaction module, and dataset of multi-person collaborative intelligent decision-making method and task distribution mechanism. This dataset is designed for the multi-station collaborative mechanism of intelligent assisted assembly. It realizes multi-person collaborative intelligent decision-making and task distribution mechanism, constructs a new efficient and adaptive guided assembly mode via assembly guidance strategies, achieves accurate expression and feedback of multi-person interaction content in complex multi-source scenarios, realizes accurate perception of component poses, and finally enables efficient and precise multi-person collaborative operation through augmented reality visual adaptive assembly guidance. 1. Standard model data of wing hoisting parts: including STEP files of skin, framework and canard shaft, as well as FBX and PLY files of canard shaft and assembly box. 2. Linemod dataset for wing canard shaft pose estimation: containing depth maps, RGB images, labels, mask maps, segmentation results (segnet_results), pose transformation matrices (transforms), model files (PLY), ground truth information file (gt.yml), camera intrinsic parameter vector file (info.yml), camera intrinsics and depth scale factor file (intrinsics.json), as well as training and testing TXT files. 3. Data of dynamic visual assembly guidance strategy and multi-person interaction module: mainly including the first 30 and last 30 pages of Word documents of source code, the instruction manual of dynamic visual assembly guidance strategy and multi-person interaction module, and FBX model files of canard shaft and assembly box. 4. Dataset of multi-person collaborative intelligent decision-making method and task distribution mechanism: mainly including OBJ files, prefab files, and the report of multi-person collaborative intelligent decision-making method and task distribution mechanism. Collection Scheme: First, measure the key geometric dimensions of the actual model, design the model structure in SolidWorks, conduct trial assembly of all parts in the assembly drawing, create a new coordinate system and set XYZ axes in the part body, then export the designed part models as STEP files. Convert the key component canard shaft into ASCII-encoded PLY format in SolidWorks, import the files into 3ds Max to convert them into FBX format data. Secondly, evenly arrange Aruco codes around the target scene for positioning, smoothly move the pan-tilt, and record data of all orientations of the target, including RGB and RGB-D images. Convert depth maps into point clouds using camera intrinsics. Calculate the initial pose based on point cloud reconstruction, filter the initial point cloud, replace the triangulated model of the initial point cloud with a high-precision model, stitch multiple segments of the dataset, generate target object masks, generate 2D and 3D bounding boxes, and finally obtain a complete dataset in standard format. Finally, import the models and prefabs into Unity. Size: Approximately 4.0 GB
