摘 要

近年来，随着经济飞速发展，我国逐步成为世界第一大汽车消费国与生产国，为世界经济的发展做出了巨大贡献。我国汽车在持续快速增长的同时，交通事故也在急剧增长，在大部分交通事故中，驾乘人员伤亡多是由于车辆碰撞引起的。翻滚碰撞虽然在所有事故中所占的比例并不高，但其致死率在交通事故中非常高。对于翻滚碰撞的研究，目前国内使用的翻滚碰撞假人存在两个主要问题：一是当前没有一款专门为翻滚碰撞试验设计的试验装置，翻滚试验使用的均是混III 50 百分位试验假人，该假人是为正面碰撞设计的；二是混III 50 百分位假人是根据美国第50百分位成年男子主要尺寸制作的，但美国人与中国人的体重及主要尺寸存在较大差异。

针对目前国内尚无翻滚实验专用假人，假人参数与中国人体实际参数不符等问题，本文设计了一套专门应用于中国五十分位人体坐姿扫描的实验台架，基于此台架确定了获取人体点云的最佳方法，通过对获得的点云数据进行三维重建及平均化处理，最终得到了基于汽车碰撞假人的中国五十分位人体几何轮廓体表模型。与以前学者的研究相比，本文工作具有以下两点创新：一是完成了基于坐姿状态下的人体三维重建；二是获得了基于翻滚碰撞下的中国50分位人体几何轮廓。在一定程度上，本文解决了建立翻滚碰撞下中国50分位人体体表模型的问题，弥补了对于汽车翻滚碰撞研究的不足。

本文完成的具体工作如下：

1. 针对坐姿状态下的中国五十分位男性人体，设计了一套人体扫描实验台架，该实验台架充分考虑了中国五十分位人体的坐姿形态，并具有旋转、定位、防遮挡功能。
2. 通过对现有的两种人体扫描方案分析，分别采用一台Kinect深度相机和两台Kinect深度相机进行扫描，并将获得点云数据进行三维重建。通过对比重建后的模型精度、尺寸及重建效率，确定了最佳扫描方案，即采用单台Kinect深度相机进行实验。
3. 利用Geomagic Studio将单台Kinect深度相机获取的点云进行点、面、形状三阶段的编辑，完成了人体的三维重建，得到多个表面较光滑、面部等细节特征保留较完整的人体体表模型。
4. 通过模板拟合与GPA算法实现了对所有模型的平均化处理，完成了建立中国五十分位人体几何轮廓的任务，结果表明，该模型表面较光滑、精度较高，能够代表中国50分位男性成年人人体轮廓。

关键词：翻滚碰撞；中国假人；三维重建；50分位

Abstract

In recent years, with the rapid development of the economy, China has gradually become the world's largest automobile consumer and producer, and has made tremendous contributions to the development of the world economy. While China's automobiles continue to grow rapidly, traffic accidents are also growing rapidly. In most traffic accidents, the casualties of drivers and passengers are mostly caused by vehicle collisions. Although the proportion of tumbling collisions in all accidents is not high, the fatality rate is very high in traffic accidents. For the study of tumbling collisions, there are two main problems in the current use of tumbling collision dummy: First, there is currently no test device specially designed for tumbling collision test. The tumbling test uses mixed III 50 percentile test leave. People, the dummy is designed for frontal collision; the second is that the mixed III 50th percentile is made according to the main size of the 50th percentile adult male in the United States, but the weight and main size of Americans and Chinese are compared. Great difference.

In view of the fact that there are no special dummy for tumbling experiments in China, and the parameters of the dummy are inconsistent with the actual parameters of the human body in China, this paper designs a set of experimental rigs specially designed for the Chinese five-seat human body posture scanning, based on which the gantry is determined. The best way to obtain the human point cloud is to obtain the Chinese five-dimensional human body contour surface model based on the car collision dummy by three-dimensional reconstruction and averaging of the obtained point cloud data. Compared with previous scholars' research, this paper has the following two innovations: one is to complete the three-dimensional reconstruction of the human body based on sitting posture; the other is to obtain the geometric contour of Chinese 50-point human body based on tumbling collision. To a certain extent, this paper solves the problem of establishing a Chinese 50-point human body surface model under rollover collision, and makes up for the shortcomings of the car roll collision research.

The specific work completed in this paper is as follows:

1) Designed a set of human body scanning experiment bench for the Chinese five-seat male body in a sitting position. The experimental bench fully takes into account the sitting posture of China's fifty-seat human body, and has rotation, positioning and anti-blocking. Features.

2) By analyzing the existing two human body scanning schemes, one Kinect depth camera and two Kinect depth cameras are used for scanning, and the point cloud data is obtained for three-dimensional reconstruction. By comparing the reconstructed model accuracy, size and reconstruction efficiency, the best scanning scheme was determined, that is, a single Kinect depth camera was used for the experiment.

3) Using Geomagic Studio to edit the point cloud acquired by a single Kinect depth camera in three stages of point, surface and shape, and complete the three-dimensional reconstruction of the human body, and obtain a plurality of human bodies with smooth surfaces and detailed features such as facial remains. Table model.

4) Through the template fitting and GPA algorithm, the averaging process of all models is realized, and the task of establishing the geometric contour of the Chinese body is completed. The results show that the model has smooth surface and high precision and can represent China 50. Human silhouette of a male adult.

Key Words：Rolling collision；Chinese dummy；3D reconstruction；50-digit

目 录

摘 要 I

Abstract II

第1章 绪论 1

1.1 研究背景及意义 1

1.1.1 研究背景 1

1.1.2 研究意义 2

1.2 研究现状 2

1.2.1 国外研究现状 2

1.2.2 国内研究现状 3

1.2.3 研究现状总结 4

1.3 研究目的及研究内容 4

1.3.1 研究目标 4

1.3.2 研究内容 4

1.3.3 研究路线 5

1.4 论文结构 6

第2章 中国50分位人体建模试验系统 7

2.1 实验台架的设计 7

2.2 扫描仪器选择 9

2.2.1 扫描仪器的介绍 9

2.2.2 扫描仪扫描原理 10

2.3 扫描系统配置 12

2.4 扫描对象的选择 12

2.4.1 志愿者的选取 12

2.4.2 人体的坐姿形态确定 13

2.5 本章小结 15

第3章 人体点云数据获取 16

3.1 点云获取理论 16

3.1.1 人体点云获取 16

3.1.2 多相机点云配准 17

3.2 点云获取实验 23

3.2.1 单台Kinect的点云获取 23

3.2.2 两台Kinect的点云获取 24

3.3 方案对比 29

3.4 本章小结 32

第4章 基于Geomagic的体表模型重建 33

4.1 体表模型生成的三个阶段 33

4.2 点云数据的优化 33

4.2.1 去除坏点 33

4.2.2 统一采样 34

4.3 人体多边形编辑 35

4.4 NURBS曲面重构 36