摘要
苹果图像检测中,对较小缺陷识别与分类时,检测时间与精度的平衡问题一直是该领域的研究重点。为实现苹果缺陷的快速高精度检测,提出了一种基于深度学习的ACE-YOLO自适应局部图像检测算法,通过深度学习缩小检测区域,利用通道注意力机制把计算机算力集中到局部检测范围以缩短检测时间,采用图像增强算法使检测细节更清晰,通过在模型中增加小目标检测层来提高检测精度。该算法利用深度学习实现局部细节检测,与常规算法相比增加了注意力机制,其检测速度提高了25%;由于引入了局部图像增强算法,并增加了小目标检测层,其在对14类苹果缺陷进行识别时,平均检测精度也由86.1%提高到95.2%。实验表明,该算法更适用于苹果缺陷的检测。
In apple image detection,the balance between detection time and accuracy of recognizing and classifying small defects has always been the research focus.In order to realize the fast and high-precision detection of apple defects,this paper proposes an ACE-YOLO adaptive local image detection algorithm based on deep learning.The algorithm adopts deep learning to reduce the detection area,uses the channel attention mechanism to focus the computer computing power on the local detection range and shorten the detection time,uses the image enhancement algorithm to make the detection details clearer,and improves the detection accuracy by adding a small target detection layer in the model.Compared with conventional algorithms,this algorithm improves the detection speed by 25%.The average detection accuracy of 14 types of apple defects is also improved from 86.1%to 95.2%,due to the introduction of local image enhancement algorithm and the addition of small target detection layer.Experiments show that the algorithm is more suitable for apple defect detection.
作者
胡天浩
高秀敏
华云松
蔡丽君
HU Tianhao;GAO Xiumin;HUA Yunsong;CAI Lijun(School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)
出处
《山东理工大学学报（自然科学版）》
CAS
2024年第1期42-47,共6页
Journal of Shandong University of Technology:Natural Science Edition
关键词
苹果缺陷
图像检测
机器学习
注意力机制
小目标检测
apple defects
image detection
machine learning
attention mechanism
small target detection
