基于改进DeepLabV3+的指针式仪表智能识别方法设计

Design of intelligent recognition scheme of pointer meter based on improved DeepLabV3+

针对现有仪表识别方法存在的诸如对表盘差异敏感、环境干扰严重以及图像质量依赖性强导致识别准确率不高的问题,提出了一种基于改进DeepLabV3+的指针式仪表智能识别算法。通过引入GhostNetV2作为主干网络进行特征提取,并添加注意力模块CBAM,有效提升了模型在仪表语义分割任务的精度;同时设计了多类仪表的示值识别算法,实现了对多类仪表的指针读数。通过在构建的指针式仪表识别数据集上对算法进行评估,结果表明,仪表智能识别算法能够适应多种仪表类型和复杂环境,识别准确率最高达99.67%,且改进的DeepLabV3+模型平均IoU达79.8%,性能优于原始模型,能够满足实际工业应用需求。

Aiming at the problems of low recognition accuracy in the existing meter recognition meth-odsdueto sensitivity to dial differences,serious environmental interference and strong image quality dependence,this paper proposes an intelligent recognition algorithm for pointer meters based on improved DeepLabV3+.By introducing GhostNetV2 as the backbone network for feature extraction,and adding the attention module CBAM,the accuracy of the model in the instrument semantic segmentation task is effectively improved.At the same time,the multi-type instrument display value recognition algorithm is designed to realize the pointer reading of multi-type instruments.The algorithm is evaluated on a constructed pointer gauge recognition dataset and the results show that the intelligent instrument recognition algorithm can adapt to a variety of instrument types and complex environments,and the recognition accuracy is as high as 99.67%.The average IoU of the improved DeepLabV3+model is 79.8%,which is better than the original model and can meet the actual industrial application requirements.