As a typical river valley city, Lanzhou's unique geographical environment and complex transportation layout pose great challenges to traffic management. This dataset is a video stream data for river valley type urban traffic intersections in complex scenarios, covering video surveillance data of multiple key traffic intersections in Lanzhou city, including video stream data under different time periods (such as morning rush hour, off peak hours), different weather conditions (such as sunny and foggy days), different traffic conditions (such as congestion, smoothness, etc.), and different lighting conditions (such as dim light, sufficient light, etc.). In terms of data processing, video data has been finely annotated and classified, including vehicle types, lanes in which vehicles are located, and other aspects. This dataset aims to provide strong data support for the development of intelligent transportation systems, traffic congestion control, and traffic accident prevention by providing rich and diverse video data of traffic intersections
The video data of the dataset comes from road monitoring at some intersections in Lanzhou city; The image data comes from video data, and at intervals of 30 seconds, select some frames from the video to convert it into images; The image annotation file is annotated using the LabelImg tool
Use the LabelImg annotation tool to annotate the image. The annotation method is as follows: select the YOLO method to annotate the vehicles in the image, and the annotation features of the vehicles are divided into lanes and vehicle types. Among them, the car models are divided into S, M, and L. S is a small vehicle, indicating vehicles that are significantly smaller than the M-type, such as tricycles, motorcycles, bicycles, etc; M is a medium-sized vehicle, such as a family car, taxi, etc; L refers to large vehicles that are significantly larger than the M-type, such as buses, trucks, buses, etc. Lane division refers to the arrangement of lanes 1, 2, 3, 4, and 5 from left to right in the selected image
Before analyzing and applying traffic surveillance video data in this article, the following data quality control and evaluation were conducted:
(1) Check the integrity and availability of video data. Firstly, check whether the video fully records the traffic situation during the specified time period, and whether the video file is damaged or lost. After inspection, the surveillance video data set in this article is complete with no missing or damaged parts.
(2) Clarity assessment. The clarity of a video is an important aspect of data quality. This article evaluates the clarity of details of vehicles and targets in videos. After inspection, it was found that in the surveillance video dataset of this article, there was fog at the intersection of Longxi Road No. 3 and Gonglin Road along Alkali Gully No. 4, and the videos were relatively blurry. The other surveillance video datasets have high resolution and no blurring phenomenon, which will not affect the subsequent recognition and analysis work.
(3) Stability check. The stability of the video is a key indicator of data quality. After inspection, the monitoring video dataset in this article does not have any issues such as shaking or tilting.
(4) Frame image annotation work verification. Check the accuracy and consistency of the annotation results to ensure their reliability. After inspection, the frame image annotations in this article are accurate and without errors
| # | number | name | type |
| 1 | 62262038 | Research on comprehensive deformation monitoring method and performance degradation prediction model of bridge in complex geological environment | National Natural Science Foundation of China |
This work is licensed under a
Creative
Commons Attribution 4.0 International License.
CSTR:11738.11.NCDC.LZJTU.DB6447.2024
10.12072/ncdc.lzjtu.db6447.2024