基于并联卷积神经网络的水果品种识别

doi:10.3969/j.issn.1004-1524.2022.11.22

浙江农业学报 ›› 2022, Vol. 34 ›› Issue (11): 2533-2541.DOI: 10.3969/j.issn.1004-1524.2022.11.22

基于并联卷积神经网络的水果品种识别

李超(), 李锋(), 黄炜嘉

江苏科技大学电子信息学院,江苏镇江 212000

收稿日期:2020-11-16 出版日期:2022-11-25 发布日期:2022-11-29
作者简介:*李锋,E-mail: lifengsl@126.com
李超(1994—),男,山东潍坊人,硕士,研究方向为机器视觉。E-mail: 1552022319@qq.com
通讯作者: 李锋
基金资助:
国家自然科学基金(61671221)

Fruit variety recognition based on parallel convolutional neural network

LI Chao(), LI Feng(), HUANG Weijia

College of Electronic Information, Jiangsu University of Science and Technology, Zhenjiang 212000, Jiangsu, China

Received:2020-11-16 Online:2022-11-25 Published:2022-11-29
Contact: LI Feng

摘要/Abstract

摘要：

为了解决传统的水果图像识别算法在特征提取上的缺陷,以及传统卷积神经网络识别率低的问题,设计了一种基于并联卷积神经网络来提取水果特征的识别方法,利用ELU激活函数替代ReLU激活函数,利用最大类间距损失函数结合传统SoftmaxWithLoss损失函数来提高对相似品种的识别准确率。选取Fruit-360数据集中的8个品种,利用边界均衡生成对抗网络(BEGAN)结合传统的数据增强方法生成大量高质量的数据集,并用其进行训练。结果表明,该模型对8个品种的平均识别准确率达98.85%,具有良好的识别效果。

关键词: 图像识别, 深度学习, 边界均衡生成对抗网络, 卷积神经网络

Abstract:

In order to solve the defects of traditional fruit image recognition algorithms in feature extraction and the low recognition accuracy of traditional convolutional neural networks, a parallel convolutional neural network was proposed to extract fruit features. ELU activation function was introduced instead of ReLU activation function in the proposed model. Besides, a combination of maximum class spacing loss function and the traditional SoftmaxWithLoss loss function was designed to improve the recognition accuracy of similar varieties. The data of 8 fruit varieties in Fruit-360 data set was selected in the present study, and enhanced by the boundary equilibrium generative adversarial network (BEGAN) combined with the traditional data augmentation to generate a large number of high-quality data for model training. It was shown that the average recognition accuracy of 8 fruit varieties reached 98.85% and exhibited good recognition effect.

Key words: image recognition, deep learning, boundary equilibrium generative adversarial network, convolution neural network

中图分类号:

S126
TP391

李超, 李锋, 黄炜嘉. 基于并联卷积神经网络的水果品种识别[J]. 浙江农业学报, 2022, 34(11): 2533-2541.

LI Chao, LI Feng, HUANG Weijia. Fruit variety recognition based on parallel convolutional neural network[J]. Acta Agriculturae Zhejiangensis, 2022, 34(11): 2533-2541.

图/表 15

图1 卷积神经网络模型示意图

Fig.1 Schematic diagram of convolution neural network model

图2 改进的网络模型结构示意图

Fig.2 Schematic diagram of improved network model

图3 边界均衡生成对抗网络的结构示意图

Fig.3 Schematic diagram of boundary equilibrium generative adversarial network

图4 判别网络结构

Fig.4 Discriminating network structure

图5 生成网络结构

Fig.5 Generate network structure

图6 实验用水果品种的图像示意

Fig.6 Images of fruit varieties used in experiment

图7 迭代5 000次(a)和10 000次(b)后的结果

Fig.7 Result after 5 000 (a) and 10 000 (b) iterations

表1 BEGAN算法与其他数据增强算法的准确率对比

Table 1 Comparison of accuracy of data enhancement with BEGAN and traditional method %

类别 Category	传统数据增强 Traditional data enhancement	基于BEGAN的数据增强 Data enhancement via BEGAN
布瑞本苹果	93.25	95.51
Apple Braeburn
红雪苹果	93.40	95.49
Apple Crimson Snow
金色苹果2	93.26	96.13
Apple Golden2
金色苹果3	93.34	95.48
Apple Golden3
樱桃1 Cherry1	93.58	95.46
樱桃2 Cherry2	94.01	96.21
Rainer樱桃	94.12	95.45
Cherry Rainer
蜡黑樱桃	93.11	95.56
Cherry Wax Black

图8 不同样本数量对准确率的影响

Fig.8 Effect of different sample quantity on accuracy

图9 不同网络模型的准确率对比

Fig.9 Comparison of accuracy of different network models

图10 不同网络模型的损失值对比

Fig.10 Comparison of loss of different network models

图11 不同损失函数的准确率对比

Fig.11 Comparison of accuracy of models with different loss functions

图12 不同损失函数的损失值对比

Fig.12 Comparison of loss of models with different loss functions

表2 引入不同损失函数的识别准确率

Table 2 Accuracy of models with different loss functions %

类别 Category	旧损失函数 Old loss function	新损失函数 New loss function
布瑞本苹果Apple Braeburn	98.30	98.31
红雪苹果Apple Crimson Snow	97.59	99.50
金色苹果2 Apple Golden2	97.15	98.90
金色苹果3 Apple Golden3	97.80	97.85
樱桃1 Cherry1	98.41	99.45
樱桃2 Cherry2	97.73	98.79
Rainer樱桃Cherry Rainer	98.10	99.15
蜡黑樱桃Cherry Wax Black	97.86	98.86

表3 不同模型的准确率对比

Table 3 Comparison of accuracy of different algorithms %

模型 Model	准确率 Accuracy
舍去a通道的单只CNN Single CNN without channel a	97.10
特征加强+BP神经网络^[5] Feature enhancement+BP neural network	98.60
形状相似的水果自动识别研究^[11] Automatic recognition of fruits with similar shape	93.00
特征提取+单只CNN^[13] Feature extraction+single CNN	95.49
本文模型Proposed model	98.85

参考文献 18

[1]	吕秋霞, 张景鸿. 基于神经网络的水果自动分类系统设计[J]. 安徽农业科学, 2009, 37(35): 17392-17394.
	LYU Q X, ZHANG J H. Design of a fruit automatic classification system based on neural network[J]. Journal of Anhui Agricultural Sciences, 2009, 37(35): 17392-17394. (in Chinese with English abstract)
[2]	陈源, 张长江. 水果自动识别的BP神经网络方法[J]. 微型机与应用, 2010, 29(22): 40-43.
	CHEN Y, ZHANG C J. Automatic recognition for fruit based on BP neural network[J]. Microcomputer & Its Applications, 2010, 29(22): 40-43. (in Chinese with English abstract)
[3]	王水平, 唐振民, 范春年, 等. 基于SVM的水果分类算法研究[J]. 武汉理工大学学报, 2010, 32(16): 44-47.
	WANG S P, TANG Z M, FAN C N, et al. Recognition algorithm of fruits based on SVM[J]. Journal of Wuhan University of Technology, 2010, 32(16): 44-47. (in Chinese with English abstract)
[4]	董立岩, 苑森淼, 刘光远, 等. 基于贝叶斯分类器的图像分类[J]. 吉林大学学报(理学版), 2007, 45(2): 249-253.
	DONG L Y, YUAN S M, LIU G Y, et al. Image classification based on Bayesian classifier[J]. Journal of Jilin University (Science Edition), 2007, 45(2): 249-253. (in Chinese with English abstract)
[5]	余肖生, 周宁, 张芳芳. 基于KNN的图像自动分类模型研究[J]. 中国图书馆学报, 2007, 33(1): 74-76.
	YU X S, ZHOU N, ZHANG F F. A KNN-based model for automatic image categorization[J]. Journal of Library Science in China, 2007, 33(1): 74-76. (in Chinese with English abstract)
[6]	DENG L M, LUAN T, MA W J. Research on maize varieties recognition system based on image processing[J]. Applied Mechanics and Materials, 2013, 397/398/399/400: 2335-2339.
[7]	王彦翔, 张艳, 杨成娅, 等. 基于深度学习的农作物病害图像识别技术进展[J]. 浙江农业学报, 2019, 31(4): 669-676. DOI
	WANG Y X, ZHANG Y, YANG C Y, et al. Advances in new nondestructive detection and identification techniques of crop diseases based on deep learning[J]. Acta Agriculturae Zhejiangensis, 2019, 31(4): 669-676. (in Chinese with English abstract) DOI
[8]	SZEGEDY C, LIU W, JIA Y Q, et al. Going deeper with convolutions[C]// 2015 IEEE Conference on Computer Vision and Pattern Recognition. Boston, MA, USA: IEEE, 2015: 1-9.
[9]	SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[EB/OL].(2004-09-04)[2020-11-16]. http://arxiv.org/pdf/1409.1556v6.pdf.
[10]	HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]// 2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, NV, USA: IEEE, 2016: 770-778.
[11]	程荣花, 马飞, 梁亚红. 形状相似水果自动化识别研究[J]. 山东农业科学, 2015, 47(8): 116-118.
	CHENG R H, MA F, LIANG Y H. Research on automatic recognition of shape similar fruits[J]. Shandong Agricultural Sciences, 2015, 47(8): 116-118. (in Chinese with English abstract)
[12]	曾平平, 李林升. 基于卷积神经网络的水果图像分类识别研究[J]. 机械设计与研究, 2019, 35(1): 23-26.
	ZENG P P, LI L S. Classification and recognition of common fruit images based on convolutional neural network[J]. Machine Design & Research, 2019, 35(1): 23-26. (in Chinese with English abstract)
[13]	徐岩, 刘林, 李中远, 等. 基于卷积神经网络的玉米品种识别[J]. 江苏农业学报, 2020, 36(1): 18-23.
	XU Y, LIU L, LI Z Y, et al. Recognition of maize varieties based on convolutional neural network[J]. Jiangsu Journal of Agricultural Sciences, 2020, 36(1): 18-23. (in Chinese with English abstract)
[14]	许伟栋, 赵忠盖. 基于卷积神经网络和支持向量机算法的马铃薯表面缺陷检测[J]. 江苏农业学报, 2018, 34(6): 1378-1385.
	XU W D, ZHAO Z G. Potato surface defects detection based on convolution neural networks and support vector machine algorithm[J]. Jiangsu Journal of Agricultural Sciences, 2018, 34(6): 1378-1385. (in Chinese with English abstract)
[15]	HUANG B, CHEN W H, WU X M, et al. High-quality face image generated with conditional boundary equilibrium generative adversarial networks[J]. Pattern Recognition Letters, 2018, 111: 72-79. DOI URL
[16]	白玉, 姜东民, 裴加军, 等. 改进的ELU卷积神经网络在SAR图像舰船检测中的应用[J]. 测绘通报, 2018(1): 125-128.
	BAI Y, JIANG D M, PEI J J, et al. Application of an improved ELU convolution neural network in the SAR image ship detection[J]. Bulletin of Surveying and Mapping, 2018(1): 125-128. (in Chinese with English abstract)
[17]	LEBEDEV V, LEMPITSKY V. Speeding up convolutional neural networks: a survey[J]. Bulletin of the Polish Academy of Sciences: Technical Sciences, 2018, 66(6): 799-811.
[18]	周飞燕, 金林鹏, 董军. 卷积神经网络研究综述[J]. 计算机学报, 2017, 40(6): 1229-1251.
	ZHOU F Y, JIN L P, DONG J. Review of convolutional neural network[J]. Chinese Journal of Computers, 2017, 40(6): 1229-1251. (in Chinese with English abstract)

基于并联卷积神经网络的水果品种识别

Fruit variety recognition based on parallel convolutional neural network

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 15

参考文献 18

相关文章 15

编辑推荐

Metrics

本文评价

[1]	吕胤春, 段恩泽, 朱一星, 郑霞, 柏宗春. 基于YOLOv8-Swin Transformer模型的翻覆肉鸭实时检测[J]. 浙江农业学报, 2025, 37(7): 1556-1566.
[2]	郑航, 冯昊栋, 薛向磊, 叶云翔, 于健麟, 俞国红. 基于实例分割的叶菜垄间导航线提取算法研究[J]. 浙江农业学报, 2025, 37(3): 701-711.
[3]	程嘉瑜, 陈妙金, 李彤, 孙奇男, 张小斌, 赵懿滢, 朱怡航, 顾清. 基于改进Faster-RCNN网络的无人机遥感影像桃树检测[J]. 浙江农业学报, 2024, 36(8): 1909-1919.
[4]	潘攀, 张建华, 郑晓明, 周国民, 胡林, 冯全, 柴秀娟. 深度学习在作物及其近缘种抗病性智能鉴定上的研究进展[J]. 浙江农业学报, 2023, 35(8): 1993-2012.
[5]	朱冬琴, 冯全, 张建华. 基于剪枝的植物病害识别方法[J]. 浙江农业学报, 2023, 35(6): 1462-1472.
[6]	白卫卫, 赵雪妮, 罗斌, 赵薇, 黄硕, 张晗. 基于YOLOv5的小麦种子发芽检测方法研究[J]. 浙江农业学报, 2023, 35(2): 445-454.
[7]	朱世松, 马婉丽, 赵理山, 郑艳梅, 郑先波, 芦碧波. 基于改进的LinkNet的苹果叶片图像分割算法[J]. 浙江农业学报, 2023, 35(1): 202-214.
[8]	陈道怀, 汪杭军. 基于改进YOLOv4的林业害虫检测[J]. 浙江农业学报, 2022, 34(6): 1306-1315.
[9]	林孝腾, 孙茜, 郑书河. 基于OpenCV的低能耗植物工厂照明系统设计与试验[J]. 浙江农业学报, 2022, 34(3): 626-635.
[10]	柴琴琴, 曾建, 张勋. 基于贝叶斯优化卷积神经网络的金线莲伪品鉴别[J]. 浙江农业学报, 2022, 34(2): 391-396.
[11]	张晴晴, 刘连忠, 宁井铭, 吴国栋, 江朝晖, 李孟杰, 李栋梁. 基于YOLOV3优化模型的复杂场景下茶树嫩芽识别[J]. 浙江农业学报, 2021, 33(9): 1740-1747.
[12]	鲍烈, 王曼韬, 刘江川, 文波, 明月. 基于卷积神经网络的小麦产量预估方法[J]. 浙江农业学报, 2020, 32(12): 2244-2252.
[13]	蒋丰千, 李旸, 余大为, 孙敏, 张恩宝. 基于Caffe卷积神经网络的大豆病害检测系统[J]. 浙江农业学报, 2019, 31(7): 1177-1183.
[14]	刘志, 贺正, 苗芳芳, 贾彪. 基于无人机的水肥一体化玉米出苗率估算方法与试验[J]. 浙江农业学报, 2019, 31(6): 977-985.
[15]	王彦翔, 张艳, 杨成娅, 孟庆龙, 尚静. 基于深度学习的农作物病害图像识别技术进展[J]. 浙江农业学报, 2019, 31(4): 669-676.