[1]李颀,陈哲豪.基于改进单次多目标检测器的果面缺陷冬枣实时检测[J].江苏农业学报,2022,38(01):119-128.[doi:doi:10.3969/j.issn.1000-4440.2022.01.014]
　LI Qi,CHEN Zhe-hao.Real-time surface defect detection of winter jujube based on improved single shot multibox detector[J].,2022,38(01):119-128.[doi:doi:10.3969/j.issn.1000-4440.2022.01.014]
点击复制

基于改进单次多目标检测器的果面缺陷冬枣实时检测()
分享到：

江苏农业学报[ISSN:1006-6977/CN:61-1281/TN]

卷:
38
期数:
2022年01期
页码:
119-128
栏目:
农业信息工程
出版日期:
2022-02-28

文章信息/Info

Title:
Real-time surface defect detection of winter jujube based on improved single shot multibox detector
作者:
李颀1; 陈哲豪2
(1.陕西科技大学电子信息与人工智能学院，陕西西安710021;2.陕西科技大学电气与控制工程学院，陕西西安710021)
Author(s):
LI Qi1; CHEN Zhe-hao2
(1.School of Electronic Information and Artificial Intelligence, Shaanxi University of Science & Technology, Xi’an 710021, China;2.School of Electrical and Control Engineering, Shaanxi University of Science & Technology, Xi’an 710021, China)
关键词:
冬枣; 果面缺陷; 实时检测; 单次多目标检测器; 多尺寸; 空间注意力模型
Keywords:
winter jujube; surface defect; real-time detection; single shot multibox detector; multi-scale; spatial attention module
分类号:
TP391.4
DOI:
doi:10.3969/j.issn.1000-4440.2022.01.014
文献标志码:
A
摘要:
为实现果面缺陷冬枣实时检测，并解决缺陷的尺寸与位置不同影响检测精度的问题，提出一种基于改进单次多目标检测器（Single shot multibox detector，SSD）的果面缺陷冬枣实时检测方法。以陕西大荔冬枣中的虫蛀、轮纹和木质化3种缺陷果和正常果为研究对象，在数据采集设备下采集实际分拣图像，然后通过数据增强由400张扩充至2 000张。改进SSD，建立MobileNetV3-SSD模型，为实时检测奠定基础；引入改进感受野块（RFB）可实现模型多尺寸提取冬枣缺陷特征的能力；用空间注意力模块（SAM）代替挤压和激励通道注意力模块（SE）增强模型定位冬枣缺陷特征的能力。试验结果表明，本研究模型在果面缺陷冬枣数据集上的表现均优于目前先进目标检测网络模型（RetinaNet和EfficientDet-D0），该模型对4类冬枣的整体检测精准性（mAP）达到91.89%，检测速度达到1 s 40.85帧。因此本研究模型较好地平衡了实时性和精准性，可应用于果面缺陷冬枣分拣流水线。
Abstract:
In order to realize the real-time surface defect detection of winter jujube and solve the problems that different sizes and positions affected the detection accuracy, a real-time surface defect detection method of winter jujube based on improved single shot multibox detector (SSD) was proposed. Three kinds of defective winter jujubes(worm, wheel-pattern and lignification) and normal winter jujubes from Dali(Shaanxi province) were taken as the research objects. The actual sorting images were collected by data acquisition equipment, and then expanded from 400 to 2 000 by data enhancement. The SSD was improved， and MobileNetV3-SSD model was established to lay the foundation for real-time detection. The introduction of improved receptive field block (RFB) could realize the ability of model to extract the defect feature of winter jujube at multiple scales. Spatial attention module (SAM) was used to replace squeeze-and-excitation (SE) block, so the ability of the model to locate the defect feature of winter jujube was enhanced. The test results showed that the performance of the proposed model on the dataset of defective winter jujube was better than the current advanced target detection network models (RetinaNet and EfficientDet-D0). The averall detection accuracy of the model for four types of winter jujube was 91.89%, and the detection speed was 40.85 frames per second. Therefore, the model established in this study can balance the real-time performance and accuracy, and can be applied to sorting pipeline of winter jujube with surface defect.

参考文献/References:

[1]王春焕,王莹,张梦辉. 大荔县冬枣产业发展建议[J].陕西林业科技,2021,49(2):97-99.
[2]侯鹏帅,刘玉乐,宋欣,等. 基于智能视觉识别的冬枣分选检测系统设计[J].中国农机化学报,2020,41(3):109-114.
[3]GENG L,XU W,ZHANG F,et al. Dried jujube classification based on a double branch deep fusion convolution neural network[J].Japanese Society for Food Science and Technology,2018,24(6):1007-1015.
[4]NTURAMBIRWE J F I, OPARA U L. Machine learning applications to non-destructive defect detection in horticultural products[J].Biosystems Engineering,2020,189:60-83.
[5]BHARGAVA A, BANSAL A. Fruits and vegetables quality evaluation using computer vision: a review[J].Journal of King Saud University-Computer and Information Sciences,2021,33(3):243-257.
[6]LU Y, LU R. Development of a multispectral structured illumination reflectance imaging (SIRI) system and its application to bruise detection of apples[J].Transactions of the ASABE,2017,60(4):1379-1389.
[7]孙世鹏,彭俊,李瑞,等. 基于近红外高光谱图像的冬枣损伤早期检测[J].食品科学,2017,38(2):301-305.
[8]JU J, ZHENG H, XU X，et al. Classification of jujube defects in small data sets based on transfer learning[DB/OL].(2021-01-30)
[2022-01-15].https://link.springer.com/article/10.1007/s00521-021-05715-2.
[9]薛勇,王立扬,张瑜,等. 基于GoogLeNet深度迁移学习的苹果缺陷检测方法[J].农业机械学报,2020,51(7):30-35.
[10]费琦琪,施杰,夏敏,等. 基于CNN的冰糖橙分级系统[J].江苏农业学报,2020,36(2):513-519.
[11]李小占,马本学,喻国威,等. 基于深度学习与图像处理的哈密瓜表面缺陷检测[J].农业工程学报,2021,37(1):223-232.
[12]LIU W, ANGUELOV D, ERHAN D, et al.SSD: single shot multibox detector[M]. Amsterdam,The Netherlands:Proceedings of the European Conference on Computer Vision,2016.
[13]徐先峰,赵万福,邹浩泉,等. 基于MobileNet-SSD的安全帽佩戴检测方法[J].计算机工程,2021,47(10):298-305,313.
[14]HOWARD A, SANDLER M, CHU G, et al. Searching for mobilenetv3[M]. Soual,Korea:International Conference on Computer Vision,2019.
[15]CHOLLET F. Xception: deep learning with depthwise separable convolutions[M]. Hawaii,United States:Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,2017.
[16]黄林生,罗耀武,杨小冬,等. 基于注意力机制和多尺度残差网络的农作物病害识别[J].农业机械学报,2021,52(10):264-271.
[17]王春山,周冀,吴华瑞,等. 改进Multi-scale ResNet的蔬菜叶部病害识别[J].农业工程学报,2020,36(20):209-217.
[18]LIU S, HUANG D, WANG Y. Receptive field block net for accurate and fast object detection[M]. Munich,Germany:Proceedings of the European Conference on Computer Vision,2018.
[19]WANDELL B, WINAWER J. Computational neuroimaging and population receptive fields[J].Trends in Cognitive Sciences,2015,19(6):349-357.
[20]HU J, SHEN L, SUN G, et al. Squeeze-and-excitation networks[M]. Salt Lake City,United States:Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,2018.
[21]高钦泉,黄炳城,刘文哲,等. 基于改进CenterNet的竹条表面缺陷检测方法[J].计算机应用,2021,41(7):1933-1938.
[22]WOO S, PARK J, LEE J, et al. CBAM: convolutional block attention module[M]. Munich,Germany:Proceedings of the European Conference on Computer Vision,2018.
[23]CHEN L, ZHANG H, XIAO J, et al. SCA-CNN: spatial and channel-wise attention in convolutional networks for image captioning[M]. Hawaii,United States:Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,2017.
[24]ZAGORUYKO S, KOMODAKIS N. Paying more attention to attention: improving the performance of convolutional neural networks via attention transfer[M]. Paris,France:International Conference on Learning Representations,2017.
[25]TSUCHIYA H, FUKUI S, IWAHORI Y, et al. A method of data augmentation for classifying road damage considering influence on classification accuracy[J].Procedia Computer Science,2019,159:1449-1458.
[26]余小东,杨孟辑,张海清,等. 基于迁移学习的农作物病虫害检测方法研究与应用[J].农业机械学报,2020,51(10):252-258.
[27]TAN M , PANG R , LE Q V. EfficientDet: scalable and efficient object detection[M]. Virtual:Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,2020.

备注/Memo

备注/Memo:
收稿日期：2021-09-08基金项目：陕西省农业科技创新工程项目
[201806117YF05NC13（1）]；陕西省科技厅农业科技攻关项目（2015NY028）；陕西科技大学博士科研启动基金项目（BJ13-15）作者简介：李颀（1973-），女，陕西西安人，博士，教授，主要研究方向为农业智能化、信息化深度学习。(E-mail)liqidq@sust.edu.cn通讯作者：陈哲豪，(E-mail)773905920@qq.com
更新日期/Last Update: 2022-03-04