基于机器视觉的机场托运行李系统行李异常滞留检测报警方法OA
Abnormal Luggage Retention Detection and Alarm Method of Airport Check-in Luggage System Based on Machine Vision
机场行李托运系统的托盘分拣机如存在分拣异常,会导致行李延误装机.针对此问题,本文提出了一种利用监控视频对其进行实时图像处理从而达到检测异常滞留行李的方法.首先对视频流进行定时抽帧处理得到正常运行下的模板图像与待处理图像,通过提取图像HOG特征向量进行余弦相似度(Cosine Similarity)计算到初步比对结果,在得到初步比对结果基础之上再次进行感知哈希算法比对(Perceptual Hash Algorithm),利用两种算法的互补性得到最终检测结果.通过为期12h对24件异常行李的检测试验证明:相较于coco数据集的YOLOv5模型与改进帧差法改方法检出率分别提升了20.8%、37.5%,并且误报率分别降低了8.3%、20.8%.通过试验证明,该方法可以部署在托盘分拣机槽口用于检测异常滞留行李以提升行李托运系统效率.
During the operation of the airport baggage handling system,the tray sorting machine needs to handle a large number of luggage,which may lead to delays in luggage loading due to sorting abnormalities.To address this issue,a method is proposed that utilizes real-time image processing of surveillance videos to detect abnormal luggage that is causing delays.First,the video stream is periodically sampled to obtain template images under normal operating conditions and the images to be processed.Image HOG feature vectors are extracted,and the Cosine Similarity is calculated to obtain a preliminary match result.Then,a perceptual hash algorithm is applied to refine the match results based on the initial comparison.The complementary nature of these two algorithms provides the final detection results.Through a 12-hour test involving 24 cases of abnormal luggage,it is demonstrated that,compared to the YOLOv5 model on the coco dataset and an improved frame difference method,the detection rate is improved by 20.8%and 37.5%respectively.Moreover,the false positive rate is reduced by 8.3%and 20.8%,respectively.This method is successfully tested and can be deployed at the tray sorting machine slots to enhance the efficiency of the baggage handling system.
陈禹州;宋洪庆;史煜青;张斌;康琳
中国民用航空总局第二研究所,成都 610041||民航成都物流技术有限公司,成都 611430
行李托运异常滞留图像处理余弦相似度感知哈希算法
checked baggageabnormal detentionimage processingcosine distanceperceptual hash algorithm
《机电工程技术》 2024 (008)
177-182 / 6
国家自然科学基金联合基金重点支持项目(U2033212);成都市区域科技创新合作项目(2023-YF11-00016-HZ)
评论