交叉口分车道流量数据集

该数据集中的为第三方采集数据。数据源验证方面，该数据集来自福田中心交通大脑平台产生的数据，具体的数据由经过严格验证的视频检测器进行采集，并要求第三方提供设备校准证书，确保数据来源可靠。该数据集保存了福田中心区2020年8月关键路口各车道的转向流量，在数据采集过程中，视频检测器以每30秒一次频率上传数据记录。为提高数据质量，数据集制备过程中计划采用多种数据质量控制手段，如对分车道数据实施车型分类逻辑校验（如大型车数量≤检测流量）；通过卡尔曼滤波算法消除视频识别误判，对“平均排队长度”与“车头时距”的物理关联性进行动态验证；对“最大排队时间”进行动态验证，对明显偏离的异常值进行强制删除；并通过人工抽验的形式确保采集的数据与道路物理特征相符。此外，采集的数据还可以与同时段其他检测器，如地磁检测器上传的流量数据进行交叉对比，确保相同时间粒度下，当量车流量的差异较小；差异过大时可通过多种数据处理手段对采集的数据进行进一步的修正，以进一步提高数据集质量。

All data within this dataset is third-party collected. For data source validation, this dataset is sourced from data generated by the Futian Central Traffic Brain Platform. Specifically, the data is collected by rigorously validated video detectors, and third-party collectors are required to submit equipment calibration certificates to guarantee reliable data provenance. This dataset contains turning flow volumes per lane at key intersections in Futian Central District during August 2020. During the data collection phase, video detectors uploaded data records at a frequency of once every 30 seconds. To enhance data quality, multiple data quality control measures will be implemented during dataset preparation: logical verification for vehicle type classification will be conducted on per-lane data (e.g., the count of large vehicles ≤ detected traffic volume); Kalman filtering algorithm will be applied to eliminate video recognition misjudgments, with dynamic verification performed on the physical correlation between "average queue length" and "headway"; dynamic verification will also be carried out for "maximum queue time", and significantly deviant outliers will be forcibly removed. Additionally, the collected data can be cross-compared with traffic flow data uploaded by other detectors operating in the same time period, such as magnetic detectors, to ensure minimal differences in equivalent vehicle traffic volume under the same time granularity. In cases of excessive differences, further corrections can be made to the collected data via multiple data processing methods to further improve the dataset's quality.

搜集汇总

背景与挑战

背景概述

该数据集是2020年8月深圳市福田中心区关键路口的分车道转向流量数据，来源于视频检测器采集，每30秒记录一次，并经过严格的质量控制（如车型校验、滤波算法和异常值处理）。数据以CSV和DOCX格式提供，总大小为211.42MB，支持交通工程研究。