摘要
为了在提高苹果树果实目标识别模型检测精度的同时减小模型体积,提出了一种基于YOLOv5s的轻量化改进模型,用于苹果树果实目标识别,并设计了相关实验。对YOLOv5s网络架构进行改进,在骨干网络中替换嵌入深度可分离卷积模块。实验结果表明,基于测试集图像,改进模型对苹果树果实目标的平均识别精度为92.4%,单张图像的识别时间为37ms,模型体积为11.2MB。与YOLOv5s相比,改进模型平均识别精度提高了0.9%,识别速率提高了16%,模型体积压缩了20%。
In order improve the detection accuracy of the apple tree fruit recognition model and reduce the model size,a lightweight approach to apple target recognition was proposed by leveraging an improved version of YOLOv5s,and relevant experiments were designed.This method improves the YOLOv5s network architecture by replacing the embedded deepwise separable convolution module in its backbone network,in order to ensure model detection accuracy while reducing the size and parameter count of the model.The experimental results show that for the test set images,the proposed improved model has an average recognition accuracy of 92.4%for apple targets,a recognition time is 37 ms for single image,and model volume is 11.2MB.Compared with the YOLOv5s model,the average recognition accuracy increases by 0.9%,the recognition speed increases by 16%,and the model volume is compressed by 20%.
作者
闫彬
邵军
YAN Bin;SHAO Jun(College of Automation and Information Engineering,Xi’an University of Technology,Xi’an 710048,China;Shaanxi Key Laboratory of Complex System Control and Intelligent Information Processing,Xi’an University of Technology,Xi’an 710048,China;College of Mechanical Engineering,Xi’an Shiyou University,Xi’an 710065,China)
出处
《实验室研究与探索》
北大核心
2025年第4期5-8,共4页
Research and Exploration In Laboratory
基金
国家自然科学基金项目(62406244)
西安石油大学教育教学改革研究项目(JGYB202321)
西安理工大学博士科研项目(103-451123011)。
关键词
目标检测
果实识别
深度学习
实验设计
object detection
fruit identification
deep learning
experimental design
