计算机工程2025，Vol.51Issue(5)：377-386,10.DOI:10.19678/j.issn.1000-3428.0069031
基于多信息融合的高速收费站拥堵检测算法
Congestion Detection Algorithm of Highway Toll Station Based on Multi-Information Fusion
摘要
Abstract
At highway toll stations,factors such as occlusion,shadows,and the depth of field significantly impact the accuracy of congestion detection.Moreover,relying solely on a single parameter for congestion detection cannot accurately reflect the actual congestion situation at toll stations.To address these challenges,this paper proposes a multistep approach.First,a target detection algorithm is employed to identify vehicles within the designated area.Spatial features extracted from the images of these vehicles are used to calculate the lane occupancy rate.Second,comprehensive vehicle flow is estimated utilizing the Deep SORT target tracking algorithm and sliding balance mechanism to mitigate false and missed detections caused by occlusion and shadowing effects.Furthermore,an optical flow-based vehicle speed estimation method is utilized to reduce congestion detection errors resulting from changes in the depth of field.Finally,through the fusion of three distinct dimensions of information(target presence,lane occupancy rate,and vehicle speed),a congestion index is obtained.The values of this index can be clustered into five categories to determine the real-time congestion state at highway toll stations.Experimental are conducted on a congested dataset collected from high-speed toll stations.The results demonstrate that the proposed multi-information fusion-based congestion detection algorithm achieves an accuracy rate of 90.4％,enabling the precise identification of traffic congestion at toll stations.关键词
多信息融合/目标检测/目标跟踪/车速估计/拥堵检测Key words
multi-information fusion/object detection/object tracking/vehicle speed estimation/congestion detection分类
信息技术与安全科学引用本文复制引用
王晓龙,江波,罗润书,安国成..基于多信息融合的高速收费站拥堵检测算法[J].计算机工程,2025,51(5):377-386,10.基金项目
国家重点研发计划(2023YFC3006700). （2023YFC3006700）
