APP下载

基于毫米波雷达的舱内儿童遗留检测系统设计和验证

2023-04-21祁淼

时代汽车 2023年8期

祁淼

摘 要:为了保护儿童避免被单独遗留在舱内,提出了基于毫米波雷达的传感器的检测方法。本方法采集毫米波多普勒效应产生的时域和频域信息,在LC-KSVD算法中加入主成分分析和随机森林的降维方法提取特征,对特征最组合。将组合的特征用SVM做分类,区分出存在和不存在儿童的场景。实验部分根据用车习惯,收集设计了正样本的采集和负样本的采集。实验表明,与同类的研究相比,本方法有更好的环境适应性可以避免相机等传统方法的局限性。

关键词:毫米波雷达 LC-KSVD算法 儿童检测 SVM分类

Design and Verification of In-cabin Child Legacy Detection System Based on Millimeter-wave Radar

Qi Miao

Abstract:In order to protect children from being left alone in the cabin, a sensor detection method based on millimeter wave radar is proposed. This method collects the time-domain and frequency-domain information generated by the millimeter wave Doppler effect, and adds the principal component analysis and the random forest dimensionality reduction method to the LC-KSVD algorithm to extract the features. The combined features are classified with SVM to distinguish the scene with and without children. In the experiment part, the collection of positive samples and negative samples are designed according to the habit of using vehicles. The experiment shows that this method has better adaptability to the environment and can avoid the limitations of traditional methods such as cameras.

Key words:millimeter wave radar, LC-KSVD algorithm, child detection, SVM classification

1 前言

汽车是许多家庭的标配,最近几年车辆设计的趋势之一是大天窗装在越来越多的车型上,2022年销量前十的车型[1]中除了五菱宏光MINIEV外都配有天窗,其中半数配置了全景天窗。如果车辆暴露在阳光下,更多的热量通过天窗传递到舱内,在密闭环境中热量聚集使舱内温度快速上升。幼儿被遗留在无人看管的车汽车里几分钟可能导致中暑和死亡。大多数父母相信自己永远不会忘记坐在后座上的孩子。现实情况是在过去的15年中美国有1000名儿童在车上因为过热去世,其中超过88%的幼儿小于3个月[2]。

常见的活体检测手段为视觉,文强[3]等人通过图像的几何形态学关系区分成年人和儿童(<6岁)的脸部特征。公妍苏[4]等人利用树莓派作为计算平台开发基于Adaptive Boosting的儿童车内遗留检测系统。但是,大多数婴儿座椅会配置遮阳帘,导致婴儿大多数特征无法被摄像头捕捉,造成漏报。……

登录APP查看全文