基于DA2-YOLOv4算法绿篱识别研究

doi:10.13733/j.jcam.issn.20955553.2022.09.017

摘要/Abstract

摘要： 针对车载式绿篱修剪机自动化修剪需要快速、准确识别绿篱的问题，提出一种DA2-YOLOv4绿篱识别算法。提出一种针对性Mosaic数据增强以获得更合理的数据扩充，使训练结果更具鲁棒性；在CSPDarknet53中引入空洞卷积得到D-CSPDarknet53，获得更大感受野，提升准确率和速度；在SPP中引入平均池化得到A-SPP,充分利用信息，使网络更具鲁棒性；删减小目标检测，获得巨大的速度提升；使用Soft-DIOU-NMS算法，提升目标重叠时的识别效果。最后制作绿篱数据集，对改进效果进行测试，并与其他算法进行对比。试验结果表明，DA2-YOLOv4算法mAP达到985%，检测速度达到83.1 FPS，较原始YOLOv4算法分别提高了8.1%和14.9 FPS，而且算法各方面性能均显著优于其他目标检测算法。DA2-YOLOv4完全满足绿篱识别要求，为绿篱修剪行业自动化提供了有力保障。

关键词: 绿篱修剪机, YOLOv4, 绿篱识别, Mosaic数据增强, 空洞卷积, Soft-DIOU-NMS

Abstract: Aiming at the problem that automatic pruning of vehicle mounted hedge trimmer needs to identify hedges quickly and accurately, a DA2-YOLOv4 hedge recognition algorithm is proposed. A targeted Mosaic data enhancement method is proposed to obtain more reasonable data expansion and make the training results more robust. D-CSPDarknet53 is obtained by introducing dilated convolution into CSPDarknet53 to obtain a larger receptive field and improve the accuracy and speed. Average pooling is introduced into SPP to obtain A-SPP, which makes full use of information to make the network more robust. The small target detection is deleted and the speed is greatly improved. The Soft-DIOU-NMS algorithm is used to improve the recognition effect when targets overlap. Finally, the hedgerow data set is made to test the improvement effect and compare with other algorithms. The experimental results show that the DA2-YOLOv4 algorithm mAP reaches 985%, and the detection speed reaches 83.1 FPS, which is 8.1% and 14.9 FPS higher than the original YOLOv4 algorithm, and the performance of the algorithm is significantly better than other target detection algorithms. The DA2-YOLOv4 fully meets the requirements of hedge identification and provides a strong guarantee for the automation of hedge pruning industry.

Key words: hedge trimmer, YOLOv4, hedge identification, Mosaic data enhancement, dilated convolution, Soft-DIOU-NMS

中图分类号:

TP391.4

韦锦, 李正强, 许恩永, 蒙艳玫, 韦和钧, 武豪. 基于DA2-YOLOv4算法绿篱识别研究[J]. 中国农机化学报, 2022, 43(9): 122-130.

Wei Jin, Li Zhengqiang, Xu Enyong, Meng Yanmei, Wei Hejun, Wu Hao.. Research on hedge recognition based on DA2-YOLOv4 algorithm[J]. Journal of Chinese Agricultural Mechanization, 2022, 43(9): 122-130.

参考文献

［1］　
Girshick R, Donahue J, Darrell T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation ［C］. 27th IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2014 Columbus OH; Institute of Electrical and Electronics Engineers, 2014.

［2］　
He K, Zhang X, Ren S, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition ［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014, 37(9): 1904.

［3］　
Girshick R. Fast R-CNN［C］. IEEE International Conference on Computer Vision, 2015 Santiago CHILE; Institute of Electrical and Electronics Engineers, 2015.

［4］　
Ren S Q, He K M, Girshick R, et al. Faster R-CNN: Towards realtime object detection with region proposal networks ［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149.

［5］　
Lin T Y, Dollar P, Girshick R, et al. Feature pyramid networks for object detection ［C］. 30th IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017 Honolulu HI; Institute of Electrical and Electronics Engineers, 2017.

［6］　
Redmon J, Divvala S, Girshick R, et al. You only look once: Unified, realtime object detection ［C］. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016 Seattle WA; Institute of Electrical and Electronics Engineers, 2016.

［7］　
Liu W, Anguelov D, Erhan D, et al. SSD: Single shot multiBox detector ［C］. 14th European Conference on Computer Vision (ECCV). 2016 Amsterdam NETHERLANDS; Springer International Publishing, 2016.

［8］　
Lin T Y, Goyal P, Girshick R, et al． Focal loss for dense object detection ［C］. 16th IEEE International Conference on Computer Vision (ICCV), 2017 Venice ITALY; Institute of Electrical and Electronics Engineers, 2017.

［9］　
Bochkovskiy A, Wang C Y, Liao H Y M. YOLOv4: Optimal speed and accuracy of object detection ［EB/OL］. https://arxiv.org/abs/2004.10934, 2020-04-23.

［10］　
郭继峰, 孙文博, 庞志奇, 等. 一种改进YOLOv4的交通标志识别算法［J/OL］. 小型微型计算机系统: 1-7［2022-06-14］. http://kns.cnki.net/kcms/detail/21.1106.TP.20210623.1130.004.html

［11］
张晴晖, 孔德肖, 李俊萩, 等. 基于逆运动学降维求解与YOLOv4的果实采摘系统研究［J］. 农业机械学报, 2021, 52(9): 15-23.

Zhang Qinghui, Kong Dexiao, Li Junqiu, et al. Design of fruit picking system based on inverse kinematics dimension reduction and YOLOv4 ［J］. Transactions of the Chinese Society for Agricultural Machinery, 2021, 52(9): 15-23.

［12］　
候瑞环, 杨喜旺, 王智超, 高佳鑫. 一种基于YOLOv4-TIA的林业害虫实时检测方法［J］. 计算机工程, 2022, 48(4): 255-261.

Hou Ruihuan, Yang Xiwang, Wang Zhichao, et al. A realtime detection method of forestry pests based on YOLOv4-TIA algorithm ［J］. Computer Engineering, 2022, 48(4): 255-261.

［13］　
蔡舒平, 孙仲鸣, 刘慧, 等. 基于改进型YOLOv4的果园障碍物实时检测方法［J］. 农业工程学报, 2021, 37(2): 36-43.

Cai Shuping, Sun Zhongming, Liu Hui, et al. Realtime detection methodology for obstacles in orchards using improved YOLOv4 ［J］. Transactions of the Chinese Society of Agricultural Engineering, 2021, 37(2): 36-43.

［14］　
Albahli S, Nida N, Irtaza A, et al. Melanoma lesion detection and segmentation using YOLOv4-DarkNet and active contour ［J］. IEEE Access, 2020, 8: 198403-198414.

［15］　
Yu Z W, Shen Y G, Shen C K, A realtime detection approach for bridge cracks based on YOLOv4-FPM ［J］. Automation in Construction, 2021, 122: 1-11.

［16］　
Simonyan K, Zisserman A. Very deep convolutional networks for largescale image recognition ［C］. International Conference on Learning Representations, 2015 San Diego; Institute of Electrical and Electronics Engineers, 2015.

［17］　
He K M, Zhang X Y, Ren S Q. Deep residual learning for image recognition ［C］. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016 Seattle WA,; Institute of Electrical and Electronics Engineers, 2016.

［18］　
Wang C Y, MarkLiao H Y, Wu Y H, et al. CSPNet: A new backbone that can enhance learning capability of CNN ［C］. IEEE Conference on Computer Vision and Pattern Recognition Workshop (CVPR Workshop), 2020 Seattle; Institute of Electrical and Electronics Engineers, 2020.

［19］　
He K, Zhang X, Ren S, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition［J］. IEEE transactions on pattern analysis and machine intelligence, 2015, 37(9): 1904-1916.

［20］　
Liu S, Qi L, Qin H, et al. Path aggregation network for instance segmentation ［C］. 2018 IEEE Conference on Computer Vision and Pattern Recognition,2018 Salt Lake City; Institute of Electrical and Electronics Engineers, 2018.

［21］　
Yu F， Koltun V. Multiscale context aggregation by dilated convolutions ［EB/OL］. http://arXiv.1511.07122-v1, 2015-11-23.

［22］　
Bodla N, Singh B, Chellappa R, et al. SoftNMS improving object detection with one line of code ［C］. 2017 IEEE International Conference on Computer Vision, 2017 Venice ITALY; Institute of Electrical and Electronics Engineers, 2017.

[1]	陈国防, , 陈兆英, , 王玉亮, , 王金星, , 范国强, , 李汉卿, . 基于数据增强深度学习的苹果花检测方法研究[J]. 中国农机化学报, 2022, 43(5): 148-155.
[2]	王权顺, 吕蕾, 黄德丰, 付思琴, 余华云. 基于改进YOLOv4算法的苹果叶部病害缺陷检测研究[J]. 中国农机化学报, 2022, 43(11): 182-187.
[3]	陈继清, 韦德鹏, 龙腾, 罗天, 王桦彬. 基于卷积神经网络的害虫分类[J]. 中国农机化学报, 2022, 43(11): 188-194.
[4]	李威, 李尚平, 潘家枫, 李凯华, 闫昱晓. 基于机器学习的中耕期甘蔗幼苗识别定位^*[J]. 中国农机化学报, 2021, 42(11): 130-137.