基于改进YOLOv4的果园柑橘检测方法研究

doi:10.16088/j.issn.1001-6600.2020111404

摘要/Abstract

摘要： 水果的自动检测是自动采摘、果园喷药、采后分拣等农业应用中的关键技术。针对果园环境中柑橘目标小、噪声多、遮挡严重等问题,本文基于YOLOv4算法提出一种改进的适用于果园环境的柑橘快速识别方法。主要改进措施包括:一是在训练阶段利用Canopy算法和K-Means++算法自动选择先验框的数目和大小;二是在YOLOv4网络中每个不同尺度特征的输出层前增加一个调整层,并采用残差网络结构和密集连接网络相结合,同时修改回归框损失函数,以便检测复杂背景下的小柑橘;三是在保证不造成较大检测精度损失的前提下,对网络中不重要的通道和网络层进行剪枝。与目前常用的YOLOv4、MLKP和 Cascade R-CNN等3种目标检测算法的对比实验结果表明,本文改进的YOLOv4算法对果园环境下不同生长期柑橘的检测平均准确率达96.04%,平均检测速度为每张图像0.06 s,均优于对比的3种主流目标检测算法。本文提出的方法可为自然条件下果园中柑橘的采摘、产量评估等应用提供技术和方法指导。

关键词: 柑橘识别, 小目标检测, 深度学习, 改进YOLOv4, 卷积神经网络

Abstract: The automatic detection of fruits is a key technology in agricultural applications such as automatic picking, orchard spraying, and post-harvesting sorting. Aiming at the problems of small citrus targets, many noises, and serious occlusion in the orchard environment, this paper proposes an improved fast identification method for citrus in the orchard environment based on the YOLOv4 algorithm. The main improvement include: one is to use the Canopy algorithm and K-Means++ algorithm to automatically select the number and size of the priori boxes in the training phase; the other is to add an adjustment layer before each output layer of different scale features in the YOLOv4 network, where the residual network structure is combined with densely connected network, and the loss function of the regression box is modified to detect small citrus in a complex background; third, on the premise of ensuring that a large amount of detection accuracy is not lost, the unimportant channels and networks in the network Layers are pruned. The experimental results of comparison with the three commonly used target detection algorithms show that the improved YOLOv4 detection method in this paper has better detection results for citrus in different growth periods in the orchard environment, with an average accuracy rate of 96.04% and a real-time detection speed of 0.06 s per image, are better than the above three mainstream target detection algorithms. The method proposed in this paper can provide technical and methodological guidance for citrus harvesting and yield evaluation in orchards under natural conditions.

Key words: citrus recognition, small target detection, deep learning, improved YOLOv4, convolutional neural networks

中图分类号:

TP391.41

陈文康, 陆声链, 刘冰浩, 李帼, 刘晓宇, 陈明. 基于改进YOLOv4的果园柑橘检测方法研究[J]. 广西师范大学学报（自然科学版）, 2021, 39(5): 134-146.

CHEN Wenkang, LU Shenglian, LIU Binghao, LI Guo, LIU Xiaoyu, CHEN Ming. Real-time Citrus Recognition under Orchard Environment by Improved YOLOv4[J]. Journal of Guangxi Normal University(Natural Science Edition), 2021, 39(5): 134-146.

参考文献

[1] 王刘坤, 祁春节. 中国柑橘主产区的区域比较优势及其影响因素研究: 基于省级面板数据的实证分析[J]. 中国农业资源与区划, 2018, 39(11): 121-128.
[2] 刘英璇, 伍锡如, 雪刚刚. 基于深度学习的道路交通标志多目标实时检测[J]. 广西师范大学学报(自然科学版), 2020, 38(2): 96-106. DOI:10.16088/j.issn.1001-6600.2020.02.011.
[3] 翟治芬, 徐哲, 周新群, 等. 基于朴素贝叶斯分类器的棉花盲椿象危害等级识别[J]. 农业工程学报, 2015, 31(1): 204-211. DOI:10.3969/j.issn.1002-6819.2015.01.028.
[4] 黄小玉, 李光林, 马驰, 等. 基于改进判别区域特征融合算法的近色背景绿色桃子识别[J]. 农业工程学报,2018, 34(23): 142-148. DOI:10.11975/j.issn.1002-6819.2018.23.017.
[5] 汪成龙, 李小昱, 武振中, 等. 基于流形学习算法的马铃薯机械损伤机器视觉检测方法[J]. 农业工程学报, 2014, 30(1): 245-252. DOI:10.3969/j.issn.1002-6819.2014.01.031.
[6] GONGAL A, AMATYA S, KARKEE M, et al. Sensors and systems for fruit detection and localization: a review[J]. Computers and Electronics in Agriculture, 2015, 116: 8-19.DOI:10.1016/j.compag.2015.05.021.
[7] LIN G C, TANG Y C, ZOU X J, et al. In-field citrus detection and localisation based on RGB-D image analysis[J]. Biosystems Engineering, 2019, 186:34-44. DOI:10.1016/j.biosystemseng.2019.06.019.
[8] 唐熔钗, 伍锡如. 基于改进YOLO-V3网络的百香果实时检测[J]. 广西师范大学学报(自然科学版), 2020, 38(6): 32-39. DOI:10.16088/j.issn.1001-6600.2020.06.004.
[9] 黄豪杰, 段先华, 黄欣辰. 基于深度学习水果检测的研究与改进[J]. 计算机工程与应用, 2020, 56(3): 127-133. DOI:10.3778/j.issn.1002-8331.1906-0384.
[10] LECUN Y, BENGIO Y, HINTON G. Deep learning[J]. Nature, 2015, 521(7553): 436-444. DOI:10.1038/nature14539.
[11] 岑冠军, 华俊达, 潘怡颖, 等. 基于深度学习芒果图像在线识别与计数方法研究[J]. 热带作物学报, 2020, 41(3): 425-432. DOI:10.3969/j.issn.1000-2561.2020.03.002.
[12] 闫建伟, 赵源, 张乐伟, 等. 改进Faster-RCNN自然环境下识别刺梨果实[J]. 农业工程学报, 2019, 35(18): 143-150. DOI:10.11975/j.issn.1002-6819.2019.18.018.
[13] KOIRALA A, WALSH K B, WANG Z, et al. Deep learning for real-time fruit detection and orchard fruit load estimation: benchmarking of ‘MangoYOLO’[J]. Precision Agriculture, 2019, 20(6): 1107-1135. DOI:10.1007/s11119-019-09642-0.
[14] 熊俊涛, 郑镇辉, 梁嘉恩, 等. 基于改进YOLO v3网络的夜间环境柑橘识别方法[J]. 农业机械学报, 2020, 51(4): 199-206. DOI:10.6041/j.issn.1000-1298.2020.04.023.
[15] KANG H W, CHEN C. Fruit detection and segmentation for apple harvesting using visual sensor in orchards[J]. Sensors, 2019, 19(20): 4599. DOI:10.3390/s19204599.
[16] FONT D, PALLEIÀ T, TRESANCHEZ M, et al. A proposal for automatic fruit harvesting by combining a low cost stereovision camera and a robotic arm[J]. Sensors, 2014, 14(7): 11557-11579. DOI:10.3390/s140711557.
[17] CHEN M Y, TANG Y C, ZOU X J, et al. Three-dimensional perception of orchard banana central stock enhanced by adaptive multi-vision technology[J]. Computers and Electronics in Agriculture, 2020, 174: 105508. DOI:10.1016/j.compag.2020.105508.
[18] HAAS T, SCHUBERT C, EICKHOFF M, et al. BubCNN: bubble detection using faster RCNN and shape regression network[J]. Chemical Engineering Science, 2020, 216: 115467. DOI:10.1016/j.ces.2019.115467.
[19] CHEN P, LI Y, ZHOU H, et al. Detection of small ship objects using anchor boxes cluster and feature pyramid network model for SAR imagery[J]. Journal of Marine Science and Engineering, 2020, 8(2): 112. DOI:10.3390/jmse8020112.
[20] TIAN Y N, YANG G D, WANG Z, et al. Apple detection during different growth stages in orchards using the improved YOLO-V3 model[J]. Computers and Electronics in Agriculture, 2019, 157: 417-426. DOI:10.1016/j.compag.2019.01. 012.
[21] WU D H, WU Q, YIN X Q, et al. Lameness detection of dairy cows based on the YOLOv3 deep learning algorithm and a relative step size characteristic vector[J]. Biosystems Engineering, 2020, 189: 150-163. DOI:10.1016/j.biosystemseng. 2019.11.017.
[22] BOCHKOVSKIY A, WANG C Y, LIAO H Y M. YOLOv4: optimal speed and accuracy of object detection[EB/OL]. (2020-04-23)[2020-11-14]. https:// arxiv.org/pdf/2004.10934v1.
[23] MISRA D. Mish: a self regularized non-monotonic neural activation function[EB/OL]. (2019-08-23)[2020-11-14]. https:// arxiv.org/pdf/1908.08681v2.
[24] ZHANG Y, ZHOU W J, WANG Y J, et al. A real-time recognition method of static gesture based on DSSD[J]. Multimedia Tools and Applications, 2020, 79(25/26): 17445-17461. DOI:10.1007/s11042-020-08725-9.
[25] 鞠默然, 罗海波, 王仲博, 等. 改进的YOLO V3算法及其在小目标检测中的应用[J]. 光学学报, 2019, 39(7): 0715004. DOI:10.3788/AOS201939.0715004.
[26] 戴加明,佟继周.基于深度残差网络的星系形态分类[J].天文学进展,2018,36(4):384-397. DOI:10.3969/j.issn.1000-8349.2018.04.03.
[27] LIU Z, LI J G, SHEN Z Q, et al. Learning efficient convolutional networks through network slimming[C] // 2017 IEEE International Conference on Computer Vision (ICCV). Los Alamitos, CA: IEEE Computer Society, 2017: 2755-2763. DOI:10.1109/ICCV.2017.298.
[28] MEHTA S S, TON C, ASUNDI S, et al. Multiple camera fruit localization using a particle filter[J]. Computers and Electronics in Agriculture, 2017, 142(Part A): 139-154. DOI:10.1016/j.compag.2017.08.007.
[29] JIANG B, WU Q, YIN X Q, et al. FLYOLOv3 deep learning for key parts of dairy cow body detection[J]. Computers and Electronics in Agriculture, 2019, 166: 104982. DOI:10.1016/j.compag.2019.104982.
[30] CAI Z W, VASCONCELOS N. Cascade R-CNN: high quality object detection and instance segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 43(5): 1483-1498. DOI:10.1109/TPAMI.2019.2956516.
[31] WANG H, WANG Q L, GAO M Q, et al. Multi-scale location-aware kernel representation for object detection[C] // 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Los Alamitos, CA: IEEE Computer Society, 2018: 1248-1257. DOI:10.1109/CVPR.2018.00136.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed