交通系统共性数据平台交通事件视频识别明细数据集

本数据集主要面向智能交通事件监测、交通安全管理及城市交通运行研究需求建设，旨在为交通事件自动识别算法评估、事件响应效率优化等提供高质量基础数据支持。数据基于深圳市福田区实际交通场景，由深圳交警部门通过第三方数据平台采集获得，时间范围为2020年8月1日至2020年8月30日，数据总量为1.84GB。数据集内容涵盖交通事件的多维识别信息，包括主事件唯一ID、事件大类与小类、视频ID、杆件ID、事件上传时间、图片与录像资源列表、算法供应商编码等。数据采集环节严格依托深圳市交委及政府开放平台的官方认证接口，确保数据源合法、权威、可追溯。为保障数据质量与可信度，数据生产过程中执行了双重质量控制策略：一方面，针对事件分类结果实施跨算法交叉验证，并将抽样记录与交警人工审核结果对比，识别准确率要求不低于95%；另一方面，对“图片列表”和“录像列表”字段采用MD5哈希算法校验文件完整性，结合FFmpeg工具检查视频帧率稳定性。此外，数据集构建了“事件-杆件-视频”三元关系图谱，并对杆件ID与GIS定位的一致性进行拓扑校验，剔除重复记录。该数据集为交通视频事件识别算法训练与评估、道路风险监测、智慧交通基础设施管理等研究与应用提供了关键数据支撑，具有良好的实用性和推广价值。This dataset is developed for the needs of intelligent traffic incident monitoring, traffic safety management and urban traffic operation research, aiming to provide high-quality basic data support for the evaluation of automatic traffic incident recognition algorithms, optimization of incident response efficiency, and other related research and applications. The data is based on actual traffic scenarios in Futian District, Shenzhen, collected by the Shenzhen Traffic Police Department through a third-party data platform, covering the time period from August 1st to August 30th, 2020, with a total data volume of 1.84 GB. The dataset covers multi-dimensional identification information of traffic incidents, including unique main event ID, major and minor event categories, video ID, pole ID, event upload time, image and video resource lists, algorithm supplier code, etc. The data collection process strictly relies on the officially certified interfaces of the Shenzhen Municipal Transportation Commission and government open platforms, ensuring that the data source is legal, authoritative and traceable. To ensure data quality and credibility, a dual quality control strategy is implemented during the data production process: on one hand, cross-algorithm cross-validation is conducted for event classification results, and sample records are compared with the manual review results of traffic police, with a required recognition accuracy of no less than 95%; on the other hand, the "image list" and "video list" fields use the MD5 hash algorithm to verify file integrity, and the FFmpeg tool is used to check the stability of video frame rates. In addition, the dataset constructs an "event-pole-video" triple relationship graph, and conducts topological verification on the consistency between pole ID and GIS positioning, eliminating duplicate records. This dataset provides key data support for research and applications such as the training and evaluation of traffic video incident recognition algorithms, road risk monitoring, and smart transportation infrastructure management, with good practicality and popularization value.

搜集汇总

背景与挑战

背景概述

该数据集面向智能交通事件监测与安全管理需求，基于深圳市福田区2020年8月的实际交通场景采集，数据量1.84GB，涵盖交通事件的多维识别信息。通过双重质量控制确保数据准确性和完整性，支持交通事件识别算法训练、道路风险监测等研究与应用。