基于YOLOv5的超声飞行时间测量算法设计  

Ultrasonic Time-of-flight Measurement Algorithm Based onDeep Learning and Artificial Intelligence

作　　者：付裕 魏艳龙 Fu Yu;Wei Yanlong(College of Computer Science and Technology,Taiyuan Normal University,Jinzhong Shanxi 030619,China)

机构地区：[1]太原师范学院计算机科学与技术学院,山西晋中030619

出　　处：《山西电子技术》2025年第2期81-84,共4页Shanxi Electronic Technology

基　　金：山西省基础研究计划(202303021211187);山西省基础研究计划(202203021212185);航空科学基金项目(202300340U0002);山西省自然科学基金项目(20210302124541)。

摘　　要：针对测量超声波飞行时间时利用互相关算法在噪声形式复杂的情况下会出现波形选取不易导致时延计算错误的问题,提出一种基于YOLOv5的超声波特征信号检测方法,引入互相关算法完成对超声特征信号时延值的计算,完成传感器的标定。训练结果表明,精确率为0.993,mAP@0.5为0.990,mAP@0.5:0.95为0.796。Aiming at the problem that the cross-correlation algorithm is used to measure the ultrasonic flight time,when the noise form is complex,the waveform selection is not easy to cause the error of time delay calculation,an ultrasonic feature signal detection method based on YOLOv5 is proposed.The cross-correlation algorithm is introduced to calculate the time delay value of the ultrasonic feature signal,and the calibration of the sensor is completed.The training results show that the accuracy is 0.993,mAP@0.5 is 0.990,and mAP@0.5:0.95 is 0.796.

关 键 词：YOLOv5 目标检测 超声测温 超声信号提取 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]