交通场景下的大型车辆右转停车检测方法研究  

Research on Right Turn Parking Detection Method for Large Vehicles in Traffic Scenarios

作　　者：王依凡 董振江 陈向东[1] WANG Yifan;DONG Zhenjiang;CHEN Xiangdong(School of Computer Science,Nanjing University of Posts and Telecommunications,Nanjing 210023,China)

机构地区：[1]南京邮电大学计算机学院,南京210023

出　　处：《计算机工程与应用》2025年第1期352-360,共9页Computer Engineering and Applications

摘　　要：随着大型车辆右转必停的新交规逐步实施,针对现有的监管方式存在效率较低的问题,提出了一种大型车辆右转停车检测方法。对YOLOv7-tiny检测模型进行改进,在ELAN模块中引入注意力机制,以提升大型车辆的检测性能。设计了一种新的大型车辆右转车道定位方法,通过自适应探测车道,提高了其在不同监控视角下的定位准确性。对大型车辆的运动轨迹进行序列数据转换,并采用结合加权平均的中值滤波算法,有效降低了原始数据的噪声。在滑动窗口损失函数中引入权重因子,进一步提升了检测方法的鲁棒性。实验结果表明,改进后的大型车辆检测模型在自制数据集和公开数据集上均取得了性能提升,大型车辆右转车道定位方法具有更强的泛化能力,并且检测到的停车时间覆盖率达到了97.4%。With the gradual implementation of the new regulation requiring large vehicles to come to a complete stop before making a right turn,a detection method for right-turning parking of large vehicles is proposed to address the inefficiency of the existing regulatory approaches.Firstly,improvements are made to the YOLOv7-tiny detection model by introducing an attention mechanism in the ELAN module to enhance the detection performance of large vehicles.Next,a new method for locating the right turn lanes of large vehicles is designed,which improves the accuracy of positioning under different monitoring perspectives by using adaptive lane detection.Finally,the motion trajectory of large vehicles is transformed into sequential data,and a weighted average median filtering algorithm is employed to effectively reduce the noise in the original data.Simultaneously,a weight factor is introduced in the sliding window loss function to further enhance the robustness of the detection method.Experimental results demonstrate that the improved detection model for large vehicles achieves performance improvements on both self-made datasets and publicly available datasets.The method for locating the right turn lanes of large vehicles exhibits stronger generalization ability,and the coverage rate of detected parking time reaches 97.4%.

关 键 词：大型车辆检测 车道定位 滤波算法 损失函数 停车时间检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]