基于改进YOLOv7的工件识别  

Workpiece Recognition Based on Improved YOLOv7

作　　者：王泽林 于晓明[3] WANG Ze-lin;YU Xiao-ming(School of Computer Science and Technology,Shenyang University of Chemical Technology,Shenyang 110142,China;Liaoning Key Laboratory of Intelligent Technology for Chemical Process Industry,Shenyang University of ChemicalTechnology,Shenyang 110142,China;Liaodong University,Dandong 118003,China)

机构地区：[1]沈阳化工大学计算机科学与技术学院,辽宁沈阳110142 [2]沈阳化工大学辽宁省化工过程工业智能化技术重点实验室,辽宁沈阳110142 [3]辽东学院,辽宁丹东118003

出　　处：《电脑与信息技术》2024年第6期12-17,共6页Computer and Information Technology

摘　　要：鉴于传统的基于模板匹配和简单机器学习算法在工件识别任务中的不足之处,提供了一种基于改进YOLOv7的工件识别方法。该方法主要聚焦于修改卷积层,将YOLOv7深层网络的标准卷积层替换为改进的卷积层,以提高网络计算速度并提高识别精度。其次,通过修改K-means聚类方法重新对数据集参数进行聚类操作,以获得更匹配工件识别的预测框。此外,引入了改进的空间金字塔池化算法,在保持感受野不变的情况下检测速度获得了提升。实验结果显示,在工件识别任务中,与传统的YOLOv7算法相比,改进后的方法提升了识别精度,并且加快了检测速度。This study proposes a workpiece recognition method based on an improved YOLOv7 to address the shortcomings of traditional template matching and simple machine learning algorithms in workpiece recognition.The method primarily focuses on modifying the convolutional layers by replacing the standard convolutional layers of the YOLOv7 deep network with improved ones,aiming to enhance network computation speed and recognition accuracy.Additionally,the dataset parameters are clustered using a modified K-means clustering method to obtain prediction boxes better suited for workpiece recognition.Moreover,an improved spatial pyramid pooling algorithm is introduced to achieve speed improvements while maintaining the receptive field unchanged.Experimental results demonstrate that the proposed method outperforms the traditional YOLOv7 algorithm in workpiece recognition tasks,achieving higher recognition accuracy and faster detection speed.

关 键 词：工件识别 目标检测 机器视觉 YOLOV7 深度学习 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]