国家体育馆观众席三维模型

为了支持国家体育馆场馆的体育展示和观众互动体验效果评估，我们研发了可以高保真复原现场观赛环境的沉浸式虚拟预演系统，支持观众席体育展示、互动、大规模声、光、影等效果的实时沉浸式预演。该系统将现场观赛的重要物理元素在虚拟环境中精确再现，对国家体育馆场馆观众席和场地进行精确建模，真实模拟灯光和材质，真实模拟音响矩阵产生的场地声场效果，同时支持视频终端、可穿戴终端、转播机位等观赛元素的模拟。
为了保证预演效果和赛时展示效果一致，需要对国家体育馆场馆进行高精度虚拟建模。其中，最主要的是观众席建模。我们依据场馆CAD图纸进行建模，制作完成了国家体育馆观众席三维模型。
本数据集主要包括国家体育馆场馆所有座椅的三维模型数据，每个座椅和真实场馆中的座椅一一对应，三维位置和真实场馆座椅位置无明显偏差。
在CAD图纸中，我们能得到场馆中每个座椅的语义信息，以及在平面方向上（x轴和y轴）的坐标信息。通过编写的自动化脚本，合并清除了场馆所有的无关元素；之后，给所有座椅分区，编上号码，编码格式为：区域号_行号_列号，例如，对于国家体育馆中101区域第一排的第一个座位，编号为101_1_1，之后的座椅依此类推。这样能保证虚拟场景中的座椅和CAD图纸中的座椅在语义上一一对应。我们使用了CAD软件中自带的导出元素坐标功能，进行了对比检测。在抽测中发现在x和y方向上的最大误差是0.5m，平均误差约0.2m。至于z轴上的高度信息，我们使用扫描模型，得出了场馆从第一排到最后一排的高度差，和每排之间的高度差，在虚拟场景中使用这个信息作为每排座椅的z轴高度。
CAD图纸获取所有座椅的位置信息之后，我们按照0.4米的尺寸将虚拟座椅布置到场景中的对应位置。 最终获得带有较精确的三维位置数据的座椅模型13156个，它们分布于46个观众席区域，共同构成了整个体育馆的观众席模型。

To support the effect evaluation of sports presentation and audience interaction experience at the National Indoor Stadium, we developed an immersive virtual rehearsal system that can restore the on-site spectator environment with high fidelity. This system supports real-time immersive rehearsal of sports presentation, interaction, large-scale sound, light, shadow and other effects in the spectator stands. The system accurately reproduces the key physical elements of on-site spectator experience in a virtual environment, precisely models the spectator stands and venue of the National Indoor Stadium, realistically simulates lighting, materials, and the venue acoustic field effect generated by the audio matrix, and supports the simulation of spectator-related elements such as video terminals, wearable terminals and broadcast camera positions.

To ensure consistency between the rehearsal effect and the on-match presentation effect, high-precision virtual modeling of the National Indoor Stadium is required, among which the spectator stand modeling is the most critical. We conducted modeling based on the venue's CAD drawings and completed the 3D model of the National Indoor Stadium's spectator stands.

This dataset mainly includes 3D model data of all seats in the National Indoor Stadium. Each seat in the dataset has a one-to-one correspondence with the actual seats in the venue, and there is no significant deviation in their 3D positions compared with the real seat positions.

From the CAD drawings, we can obtain the semantic information of each seat in the venue, as well as their coordinate information in the planar direction (x-axis and y-axis). We used custom-developed automated scripts to merge and remove all irrelevant elements in the venue. Subsequently, all seats were partitioned and numbered in the format of: region_number_row_number_column_number. For example, the first seat in the first row of area 101 in the National Indoor Stadium is numbered 101_1_1, and the rest seats follow the same rule. This ensures that the seats in the virtual scene have a one-to-one semantic correspondence with the seats in the CAD drawings. We used the built-in element coordinate export function of CAD software for comparative verification. The spot check results showed that the maximum error in the x and y directions was 0.5m, with an average error of approximately 0.2m. As for the height information on the z-axis, we used the scanned model to obtain the height difference from the first row to the last row of the venue, as well as the height difference between each row, and applied this information as the z-axis height of each row of seats in the virtual scene.

After obtaining the position information of all seats from the CAD drawings, we placed the virtual seats at their corresponding positions in the scene according to the size of 0.4 meters. Finally, we obtained 13,156 seat models with relatively accurate 3D position data, which are distributed across 46 spectator stand areas and jointly constitute the spectator stand model of the entire stadium.

搜集汇总

背景与挑战

背景概述

该数据集包含国家体育馆观众席的精确三维模型，基于CAD图纸构建了13156个座椅的三维位置数据，分布于46个区域，用于支持沉浸式虚拟预演系统的体育展示和观众互动体验评估。数据由中国传媒大学研发，属于计算机图形学应用领域。