慢行交通速度评估数据集 (NTSED)
本期开放慢行交通速度评估数据集，开放的数据发布在OpenITS官方网站（http://www.openits.cn/），可供所有人免费下载使用，这些数据仅限用于非商业用途，且基于开放数据的研究成果必须标注数据由OpenITS提供。
1. 数据集概述
慢行交通速度评估数据集用于测试验证基于路侧视频的慢行交通速度计算方法，是实验室自建数据集。数据集具体情况可参考以下论文：
李熙莹, 陈丽娟. 一种基于融合网络的慢行交通速度计算方法[J]. 交通运输系统工程与信息, 2022, 22(04): 186-193.
2. 数据内容
慢行交通速度评估数据集用于测试验证基于路侧视频的慢行交通速度计算方法。因目前缺乏专用于验证慢行交通速度计算方法的数据集，因此实验室自建了双视角协同数据集。数据集中有两种视频，分别是高空视频与路侧视频。高空视频数据采用无人机悬停俯拍，用于作为速度计算参考，共有8组数据，其中步行数据4组，组均 3844 帧，骑行数据 4 组，组均 2343 帧，包含可计算速度数据750多条，包括不同间隔帧的参考速度等，场景如图1（a）所示。路侧视频采用固定安装的监控摄像机拍摄，用于算法测试，共有12段视频数据，其中步行数据6段，骑行数据 6段，场景如图1（b）所示。数据标签中包含可计算速度数据750多条，包括不同间隔帧的参考速度等，也提供目标在图像上的位置坐标，以及实际拍摄场景的真实距离等数据。每组数据的标签都与单人标签一致。
在数据集中，采用不同的交通方式，包含步行与骑行。同时，数据集中增加并排行走/骑行、穿插行走/骑行，使得目标行动轨迹不规则、速度变化范围大，情况多样化，更加符合实际情况，以更好地验证算法的有效性、普适性。数据集内容如表1所示。
3. 数据文件
包含4个压缩包文件：
Overhead Video.zip：由该文件进行解压缩
Overhead Video.z01：压缩分卷文件
Roadside Video.zip：由该文件进行解压缩
Roadside Video.z01：压缩分卷文件
4. 数据文件描述
Overhead Video.zip，Overhead Video.z01：包含12段原始视频，8组速度数据、目标图像位置，实际拍摄场景的真实距离，视频均为.mp4格式，图片均为.jpg格式。
Roadside Video.zip，Roadside Video.z01：包含12段原始视频，实际拍摄场景的真实距离，视频均为.mp4格式。
Non-motorized Traffic Speed Evaluation Dataset (NTSED)
1. Dataset Overview
The non-motorized traffic speed evaluation dataset (NTSED) is used to test and validate the roadside video-based slow traffic speed calculation method, and is a self-built dataset in the laboratory. The details of the dataset are described in the paper:
Xiying Li, Lijuan Chen. A Method for Calculating Speed of Non-motorized Transportation Based on Fusion Network [J]. Transportation System Engineering and Information, 2022, 22(04): 186-193.
2. Dataset Description
The NTSED is used to test and validate the roadside video-based non-motorized traffic speed calculation method. Since there is a lack of datasets dedicated to validate the non-motorized traffic speed calculation method, we constructed this dual-view collaborative dataset. In the dataset, there are two types of videos, namely overhead video and roadside video. The overhead video data are taken by UAV hovering and swooping, and used as a reference for speed calculation. There are 8 groups of data, including 4 groups of walking data with 3844 frames and 4 groups of cycling data with 2343 frames, containing more than 750 items of calculable speed data, including reference speeds of different interval frames, etc, and the scene is shown in Figure 1(a). The roadside video is taken by a fixed installation surveillance camera for algorithm testing, with 12 segments of raw video, which are 6 walking data and 6 cycling data, and the scene is shown in Figure 1(b). The data labels contain more than 750 items of computable velocity data, including reference velocities of different interval frames, etc., and also provide data such as the location coordinates of the target on the image, and the real distance of the actual filmed scene. The labels of each data set are consistent with the single label.
In the dataset, different modes of transportation are used, including walking and cycling. At the same time, side-by-side walking/interspersed walking and side-by-side riding/interspersed cycling are also designed in the dataset, making the target action trajectory irregular, speed varying in a wide range, and the situation variable, which is more realistic to better verify the effectiveness and universality of the algorithm. The dataset content is shown in Table 1.
3. Data files
There are 4 compressed packages:
Overhead video.zip
Overhead video.z01
Roadside video.zip
Roadside video.z01
4. Data files description
Overhead video.zip, Overhead video.z01: contains 12 segments of raw video, 8 sets of speed data, target image position, real distance of actual shooting scene, all videos are in ‘.mp4’ format, all images are in ‘.jpg’ format.
Roadside video.zip, Roadside video.z01: contains 12 segments of raw video, the real distance of the actual shooting scene, the videos are all in ‘.mp4’ format.
数据提供单位
本数据由中山大学提供。
The dataset is provided by：Sun Yat-Sen University.
引文格式
李熙莹，OpenITS联盟. OpenData V14.0-慢行交通速度评估数据集 (NTSED). http://www.openits.cn/openData4/823.jhtml (2022). Accessed: 2022-XX-XX.（时间信息请按实际情况更改）
李熙莹,陈丽娟.一种基于融合网络的慢行交通速度计算方法[J].交通运输系统工程与信息,2022,22(04):186-193.DOI:10.16097/j.cnki.1009-6744.2022.04.021.
Citation Reference
Li, X. &OpenITS. OpenData V14.0-Non-motorized Traffic Speed Evaluation Dataset (NTSED). http://www.openits.cn/openData4/823.jhtml (2022). Accessed: 2022-XX-XX.（Please change the accessed data accordingly.）
Xiying Li, Lijuan Chen. A Method for Calculating Speed of Non-motorized Transportation Based on Fusion Network [J]. Journal of Transportation Systems Engineering and Information Technology, 2022, 22(04): 186-193. DOI: 10.16097/j.cnki.1009-6744.2022.04.021.
相关联系人：
广东省智能交通系统重点实验室 吴荔非 openits@126.com
If you have any problems, please contacts：
Lifei Wu, Sun Yat-Sen University, openits@126.com
注 ：下载数据后解压时请使用除winRAR以外的解压工具进行解压