The invention discloses a light neural network model edge end cross identification-based method and system for a vehicle to give way to pedestrians, and belongs to the technical field of identification of vehicle to give way to pedestrians. Comprising the following steps: S1, a vehicle-mounted pavement video collector of a bus video collection system collects three groups of image data of a zebra crossing-free image D1, a zebra crossing-free pedestrian D2 and a zebra crossing-free pedestrian D3 in a driving process; s2, carrying out image preprocessing on the collected image data; s3, training a zebra crossing identification model by using the negative sample (D2 + D3) and the positive sample D1, and identifying a pedestrian on the zebra crossing by using the positive sample D3 and the negative sample D2; s4, when the vehicle-mounted camera does not detect the zebra crossing on the road, only the zebra crossing identification model is kept for monitoring, and when the zebra crossing on the road surface is detected; the monitoring data is selectively collected, the multi-dimensional interaction state of the vehicle and the road condition of the target traffic light area is considered, and the behavior of the vehicle to give way to pedestrians is quantified in a scoring mode.
本发明公开了轻型神经网络模型边缘端交叉识别的车辆礼让行人方法、系统,属于识别车辆礼让行人技术领域,包括以下步骤:S1:公交视频采集系统的车载路面视频采集器采集行车过程中无斑马线图片D1、斑马线上无行人D2、斑马线上有行人D3的三组图片数据;S2:采集后的图片数据经过图像预处理;S3:用负样本(D2+D3)、正样本D1训练斑马线识别模型,用正样本D3、负样本D2斑马线上行人识别模型;S4:当车载中的摄像头没有检测到道路上有斑马线时,只保留斑马线识别模型监听,当检测到路面有斑马线时;有选择地采集监听数据考虑车辆与目标红绿灯区域路况的多维度交互状态,通过打分的方式,量化车辆礼让行人行为。
Method and system for allowing vehicles to give way to pedestrians through light neural network model edge end cross recognition
轻型神经网络模型边缘端交叉识别的车辆礼让行人方法、系统
2022-11-29
Patent
Electronic Resource
Chinese
Traffic light scheduling for pedestrians and vehicles
IEEE | 2017
|Roadside system for prompting passing vehicles when pedestrians cross street
European Patent Office | 2021
|