The invention belongs to the technical field of traffic flow detection, and provides a traffic flow monitoring method based on sound source localization, and the method comprises the steps: S1, sound pickup equipment is installed; s2, adding an acoustic feature enhancement band; s3, constructing a multi-task training model, and training to obtain a statistical feature scatter sequence set; s4, collecting audio data in real time, and generating a scatter time sequence of sound peaks; s5, comparing the statistical characteristic scatter point sequence set through the characterization characteristic parameters of the peak scatter points, and screening vehicle sound peak characteristics; s6, counting effective vehicles to obtain real-time traffic flow data; and S7, optimizing the multi-task training model, and updating the statistical feature scatter sequence set. According to the method, through application of the acoustic feature enhancement band and the deep learning model, the accuracy of vehicle counting and classification is greatly improved, and the method has the characteristics of low cost and easy deployment.
本发明属于交通流量检测技术领域,提出了一种基于声源定位的车流量监测方法,该方法包括:步骤S1、拾音设备安装;步骤S2、增设声学特征增强带;步骤S3、多任务训练模型构建,训练得到统计特征散点序列集;步骤S4、实时采集音频数据,生成声音尖峰的散点时间序列;步骤S5、通过尖峰散点的表征特性参数比对统计特征散点序列集,筛选车辆声音尖峰特征;步骤S6、有效车辆计数,得到实时车流量数据;步骤S7、多任务训练模型优化,统计特征散点序列集更新。本方法通过声学特征增强带和深度学习模型的应用,大幅提高车辆计数和分类的准确度,且具有低成本与易部署的特点。
Traffic flow monitoring method based on sound source localization
一种基于声源定位的车流量监测方法
2024-09-03
Patent
Electronic Resource
Chinese
Moving sound source localization based on triangulation method
Online Contents | 2016
|Moving sound source localization based on triangulation method
Online Contents | 2016
|European Patent Office | 2018
|Vehicle-based sound source localization method and device, medium and equipment
European Patent Office | 2024
|