基于元学习的小样本指针式仪表检测方法  

Few-shot pointer meters detection method based on meta-learning

作　　者：孙前来[1] 林绍杭 刘东峰 宋晓阳 刘佳耀 刘瑞珍[1] SUN Qianlai;LIN Shaohang;LIU Dongfeng;SONG Xiaoyang;LIU Jiayao;LIU Ruizhen(School of Electronic Information Engineering,Taiyuan University of Science and Technology,Taiyuan Shanxi 030024,China)

机构地区：[1]太原科技大学电子信息工程学院,山西太原030024

出　　处：《图学学报》2025年第1期81-93,共13页Journal of Graphics

基　　金：山西省重点研发计划项目(202102020101005);山西省高等学校科技创新项目(2023L185);来晋工作优秀博士奖励资金(20222088);太原科技大学研究生教育创新项目(SY2023016)。

摘　　要：仪表定位精度是保证指针式仪表示数准确识别的前提。复杂工业场景下仪表样本难以采集,小样本情况下,现有指针式仪表定位检测方法存在检测精度低、实时性差的问题。为此,提出了基于元学习的小样本指针式仪表检测方法Sparse-Meta-DETR。并以Meta-DETR为目标检测基线模型,采用元学习训练策略,在元训练阶段构建多个小样本任务训练Sparse-Meta-DETR模型,增强特征相关聚合模块对特征空间中支持集和查询集类别的余弦距离的度量能力,使模型能够在元测试阶段小样本任务中识别图像包含的类别,快速适应新类小样本任务,检测复杂工业场景图像中包含的指针式仪表;引入轻量级主干网络Efficientnet b1作为特征提取器,减少模型的计算复杂度和参数量;设计评分网络对查询特征稀疏采样,构建稀疏化遮罩选取前景特征,引导Transformer编/解码器对前景特征进行处理,进一步减少计算量并提高检测精度。使用Sparse-Meta-DETR模型,20-shot时指针式仪表定位检测精度指标AP^(50)和AP^(75)分别达到了94.2%和87.5%,10-shot时的AP^(50)达到了91.1%;相较于最初的基线模型,改进模型的时间复杂度下降了74.5%。实验结果表明,Sparse-Meta-DETR不仅能够保证样本匮乏时仪表定位的精度,还可以有效地提高仪表定位的实时性,其整体性能优于Meta-RCNN等小样本深度学习算法。The accuracy of meters location is a critical factor in ensuring the accuracy of meters recognition.However,it is challenging to collect instrument data in complex industrial scenarios,existing pointer instrument detection methods exhibit low detection accuracy and poor real-time performance in few-shot situations.For this reason,the Sparse-Meta-DETR method was proposed for few-shot pointer meter detection based on meta-learning.Inspired by the object detection model Meta-DETR,this method adopted the meta-learning strategy.During the meta-training stage,few-shot tasks were created to train the Sparse-Meta-DETR model,enhancing metrics ability of the correlational aggregation module for support set and query set classes in the feature space.This enabled the model to recognize classes present in images during the few-shot training stage with few-shot tasks,quickly adapt to few-shot tasks with novel classes,detect pointer meters in complex industrial scenarios.A lightweight backbone network,Efficientnet b1,was introduced as the feature extractor to reduce the computational complexity and parameter of the model,thereby improving the detection speed.Simultaneously,a scoring network was designed as a token sparsification sampler,creating a sparsification mask to select foreground features from query features.This guided the Transformer encoders and decoders to focus on foreground features,thereby reducing computational complexity of few-shot training stage and improving detection accuracy.The Sparse-Meta-DETR model achieved an AP^(50) of 94.2%and an AP^(75) of 87.5%in 20-shot task,and an AP^(50) of 91.1%in 10-shot tasks.Compared to the baseline model,the improved model reduced time complexity by 74.5%.Experimental results demonstrated that the Sparse-Meta-DETR can not only effectively ensure the accuracy of pointer meter positioning detection but also improve the real-time performance in the case of few-shot.Its overall performance surpassed other few-shot deep-learning algorithms such as Meta RCNN.

关 键 词：指针式仪表 元学习 小样本 目标检测 稀疏采样 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP18[自动化与计算机技术—计算机科学与技术]