摘要
为了对铁路电气设备进行智能检测,并提升检测精度,采用了You Only Look Once version 4算法,并从特征获取、全局信息获取和损失函数三个角度对其进行初步改进。基于提升模型检测速度的目的,又引入了K均值算法及深度可分离卷积。经过测试发现,在实际应用中,检测模型的准确率和召回率最大值分别为93.83%和94.61%,耗时和内存占用率最小值分别为22.61 ms和5.8%。所设计的智能检测模型具有良好的检测精度和速度,能够较好地对现实中的铁路电气设备进行检测。
In order to intelligently detect railway electrical equipment and improve detection accuracy,the You Only Look Once version 4 algorithm was adopted,and preliminary improvements were made from three perspectives:feature acquisition,global information acquisition,and loss function.In order to improve the speed of model detection,K-means clustering algorithm and depthwise separable convolution were introduced.After testing,it was found that in practical applications,the maximum accuracy and recall of the detection model were 93.83%and 94.61%,respectively,while the minimum time consumption and memory usage were 22.61ms and 5.8%,respectively.The designed intelligent detection model has good detection accuracy and speed,and can effectively detect railway electrical equipment in reality.
作者
娄刘娟
LOU liujuan(Xi’an Railway Vocational&Technical Institute,Xi’an 710026,China)
出处
《自动化与仪器仪表》
2025年第7期33-37,共5页
Automation & Instrumentation
基金
《BIM技术在接触网腕臂预配中的应用研究》(XTZY20J21)。
