弱光条件下猪舍清洗目标检测

doi:10.3969/j.issn.1004-1524.20221447

浙江农业学报 ›› 2023, Vol. 35 ›› Issue (9): 2240-2249.DOI: 10.3969/j.issn.1004-1524.20221447

弱光条件下猪舍清洗目标检测

李颀(), 李煜哲()

陕西科技大学电子信息与人工智能学院,陕西西安 710021

收稿日期:2022-10-10 出版日期:2023-09-25 发布日期:2023-10-09
作者简介:*李煜哲,E-mail: 1210412528@qq.com
李颀(1969—),女,陕西西安人,博士,教授,主要从事工业自动化与智能控制等方面的教学与科研工作。E-mail: liqidq@sust.edu.cn
通讯作者: 李煜哲
基金资助:
陕西省科技厅项目(S2023-YF-YBNY-0232)

Detection of piggery cleaning target under low-light condition

LI Qi(), LI Yuzhe()

College of Electronic Information and Artificial Intelligence, Shaanxi University of Science and Technology, Xi’an 710021, China

Received:2022-10-10 Online:2023-09-25 Published:2023-10-09
Contact: LI Yuzhe

摘要/Abstract

摘要：

无人工干预的猪舍清洗可以有效预防猪只感染疾病。针对猪舍场景中光线较弱、特征不明显、存在相互遮挡情况导致对清洗目标的检测准确率低的问题,提出一种改进的YOLOv5算法。在预处理阶段,利用基于双边滤波的Retinex算法提高在弱光条件下对猪舍清洗目标的检测能力;在清洗目标检测过程中,通过在YOLOv5网络的backbone中引入CBAM(convolutional block attention module)注意力机制,使网络学习到猪舍清洗目标更多的有效特征信息;对网络中边框回归损失函数进行改进,采用DIoU-NMS算法筛选出猪舍清洗目标的候选框,提高在部分遮挡情况下清洗目标的检测精度。实验结果表明:在猪舍清洗目标的测试集上改进后的YOLOv5目标检测算法较基准算法准确率提高7.3百分点,召回率提高7.6百分点,平均准确度提高7.1百分点,在弱光条件和目标遮挡情况下鲁棒性更高。研究结果为畜牧智能清洗设备的研发提供了基础。

关键词: 猪舍, 清洗目标, 弱光, 注意力机制, YOLOv5算法

Abstract:

The cleaning of piggery without manual intervention can effectively prevent pigs from contracting diseases. Aiming at the problems of weak light, inconspicuous features, and mutual occlusion in the piggery scene, the low detection accuracy of cleaning targets, an improved YOLOv5 algorithm was proposed. In the preconditioning stage, the Retinex algorithm based on bilateral filtering was used to improve the detection ability of the piggery cleaning target under low light conditions; In the cleaning target detection process, the CBAM (convolutional block attention module) attention mechanism was introduced into the backbone of the YOLOv5 network to enable the network to learn more effective feature information of the cleaning target in the piggery; The frame regression loss function in the network was improved, and the DIoU-NMS algorithm was used to filter out the candidate frame of the cleaning target in the piggery, which improved the detection accuracy of the cleaning target in the case of partial occlusion. The experimental results showed that the improved YOLOv5 target detection algorithm improved the accuracy rate by 7.3 percentage point, the recall rate by 7.6 percentage point, and the average accuracy by 7.1 percentage point compared with the benchmark algorithm on the test set of piggery cleaning targets. The robustness was higher in the case of target occlusion. The result provided a basis for the research and development of animal husbandry intelligent cleaning equipment.

Key words: piggery, cleaning target, low-light conditions, attention mechanism, YOLOv5 algorithm

中图分类号:

S237

李颀, 李煜哲. 弱光条件下猪舍清洗目标检测[J]. 浙江农业学报, 2023, 35(9): 2240-2249.

LI Qi, LI Yuzhe. Detection of piggery cleaning target under low-light condition[J]. Acta Agriculturae Zhejiangensis, 2023, 35(9): 2240-2249.

图/表 11

图1 特征金字塔(FPN)+金字塔注意网络(PAN)结构

Fig.1 The structure of feature pyramide network (FPN)+pyramid attention network (PAN)

图2 猪舍清洗目标

Fig.2 Piggery cleaning target

图3 本文改进的YOLOv5算法框架 Conv,卷积;BN,归一化;CBL,卷积块;CSPNet,跨阶段局部网络;SPP,空间金字塔池化;Concat,张量拼接;Leaky relu,激活函数;Res unit,残差组件;CBAM,轻量级的卷积注意力模块。

Fig.3 Algorithm framework of the improved YOLOv5 network Conv, Convolution; BN, Batch normalization; CBL, Convolution block; CSPNet, Cross stage partial networks; SPP, Spatial pyramid pooling; Concat, Concatenates tensors; Leaky relu, Activation function; Res unit, Residual component; CBAM, Convolutional block attention module.

图4 基于双边滤波的Retinex猪舍图像增强算法

Fig.4 Retinex piggery image enhancement algorithm based on bilateral filtering

图5 猪舍弱光图像增强 a,图像增强前;b,图像增强后。

Fig.5 Low-light image enhancement in the piggery a, Before image enhancement; b, After image enhancement.

图6 CBAM模块基本结构

Fig.6 The structure of CBMA model

图7 模型的损失函数变化曲线

Fig.7 Loss function change curve of model

表1 消融实验结果

Table 1 Ablation experiments results

序号 Serial number	图像增强 Image enhancement	加入CBAM Input CBAM	修改损失函数 Modify loss function	mAP/%
1	×	×	×	89.7
2	√	×	×	90.9
3	×	√	×	93.7
4	×	×	√	91.5
5	√	√	×	94.9
6	√	×	√	91.7
7	×	√	√	95.8
8	√	√	√	96.8

表2 不同算法的猪舍清洗目标检测结果对比

Table 2 Comparison of piggy cleaning target detection results with different algorithms

算法	P/%	R/%	mAP/%	FPS
Algorithm
Faster R-CNN	92.0	89.8	94.1	13
SSD	84.3	78.6	82.9	28
YOLOv5	89.1	84.2	89.7	30
Improved YOLOv5	96.4	91.8	96.8	30

图8 预处理前后清洗目标检测结果

Fig.8 Target test results before and after pretreatment

图9 不同算法遮挡情况清洗目标检测结果 a,无遮挡;b,部分遮挡且背景复杂;c,部分遮挡且背景简单。

Fig.9 Clean target detection results for different algorithms a, No occlusion; b, Partial occlusion and complex background; c, Partial occlusion and simple background.

参考文献 13

[1]	黄静, 张健. 一种基于改进SSD网络的猪个体目标检测方法研究[J]. 软件工程, 2022, 25(8): 25-29.
	HUANG J, ZHANG J. Research on a target detection algorithm of individual pig based on improved SSD network[J]. Software Engineering, 2022, 25(8): 25-29. (in Chinese with English abstract)
[2]	GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]// Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition. New York: ACM, 2014: 580-587.
[3]	HE K M, ZHANG X Y, REN S Q, 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.
[4]	GIRSHICK R. Fast R-CNN[C]// 2015 IEEE International Conference on Computer Vision (ICCV). December 07-13, 2015, Santiago, Chile. IEEE, 2016: 1440-1448.
[5]	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.
[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). June 27-30, 2016, Las Vegas, NV, USA. IEEE, 2016: 779-788.
[7]	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 S P, SUN Z M, LIU H, 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. (in Chinese with English abstract)
[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.
	CHEN D H, WANG H J. Detection of forest pests based on improved YOLOv4[J]. Acta Agriculturae Zhejiangensis, 2022, 34(6): 1306-1315. (in Chinese with English abstract)
[9]	LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot MultiBox detector[M]//Computer Vision-ECCV 2016. Cham: Springer International Publishing, 2016: 21-37.
[10]	WEI C, WANG W J, YANG W H, et al. Deep retinex decomposition for low-light enhancement[EB/OL]. 2018: arXiv: 1808.04560. https://arxiv.org/abs/1808.04560.
[11]	WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module[M]// Computer Vision-ECCV 2018. Cham: Springer International Publishing, 2018: 3-19.
[12]	陈世权, 王从庆, 周勇军. 一种基于YOLOv5s和图像融合的行人检测方法[J]. 电光与控制, 2022, 29(7): 96-101.
	CHEN S Q, WANG C Q, ZHOU Y J. A pedestrian detection method based on YOLOv5s and image fusion[J]. Electronics Optics & Control, 2022, 29(7): 96-101. (in Chinese with English abstract)
[13]	王杰, 陈黎卿, 黄莉莉, 等. 基于Retinex的弱光条件下车道线识别方法[J]. 计算机与数字工程, 2019, 47(2): 451-456.
	WANG J, CHEN L Q, HUANG L L, et al. Lane recognition method in weak light condition based on retinex[J]. Computer & Digital Engineering, 2019, 47(2): 451-456. (in Chinese with English abstract)

编辑推荐

Metrics

阅读次数

全文

482

HTML			PDF

最新录用	在线预览	正式出版	最新录用	在线预览	正式出版
0	0	4	0	0	478

来源	本网站	其他网站

次数	328	154
比例	68%	32%

摘要

119

最新录用	在线预览	正式出版

0	0	119

	来源	本网站

	次数	119
	比例	100%

[1]	丁东霞, 李能慧, 李静, 唐超男, 王成, 牛天航, 杨滟, 杨海涛, 颉建明. 外源褪黑素对低温弱光胁迫下辣椒叶绿素荧光和抗氧化系统的影响[J]. 浙江农业学报, 2022, 34(9): 1935-1944.
[2]	唐卫东, 刘振文, 刘冬生, 胡雪华. 低温弱光胁迫对设施黄瓜叶片面积与干物质量的影响[J]. 浙江农业学报, 2022, 34(3): 517-524.
[3]	周品志, 裴悦琨, 魏冉, 张永飞, 谷宇. 基于YOLOV4模型的果园樱桃实时检测研究[J]. 浙江农业学报, 2022, 34(11): 2522-2532.
[4]	张宁, 吴华瑞, 韩笑, 缪祎晟. 基于多尺度和注意力机制的番茄病害识别方法[J]. 浙江农业学报, 2021, 33(7): 1329-1338.
[5]	苟秉调, 段盼盼, 杨楠, 赵淑芳, 王永富, 张高原, 魏兵强. 低温弱光胁迫下辣椒苗期光合相关指标的杂种优势[J]. 浙江农业学报, 2021, 33(3): 429-436.
[6]	李莉, 田士林, 姜俊, 宋丽. 一氧化氮对弱光胁迫下苗期及坐果初期辣椒生长和抗性相关指标的影响[J]. 浙江农业学报, 2019, 31(12): 2036-2042.
[7]	高青海，王亚坤，郭远远.. 薄皮甜瓜种质资源苗期耐低温弱光鉴定及形态指标选择[J]. 浙江农业学报, 2016, 28(8): 1360-.
[8]	吕晓菡;柴伟国. 低温弱光下不同起源地辣椒幼苗光合特性的比较研究[J]. , 2014, 26(1): 0-53.
[9]	周雅倩;陆国权. 主成分分析法对观赏甘薯品种耐弱光性综合评价[J]. , 2013, 25(6): 0-1201.
[10]	陈绍孟;路伏增;*;华卫东;徐如海;褚晓红;戴丽荷;胡锦平;宋维龙. 舍内喷洒中草药制剂预防断奶仔猪腹泻的试验效果[J]. , 2013, 25(6): 0-1227.
[11]	丁小涛;金海军;张红梅;余纪柱*. 6-BA缓解黄瓜弱光胁迫的浓度研究[J]. , 2010, 22(6): 769-773.
[12]	张雅;何勇;朱祝军;王永传. 低温弱光对茄子嫁接苗和自根苗抗氧化酶系统的影响[J]. , 2008, 20(6): 0-431.
[13]	刘永华;吴晓花;李国景;汪宝根;鲁忠富. 低温弱光对生态型瓠瓜幼苗生长和生理生化特性的影响[J]. , 2006, 18(6): 0-424.
[14]	李国景;刘永华;吴晓花;汪宝根;鲁忠富. 长豇豆品种耐低温弱光性和叶绿素荧光参数等的关系[J]. , 2005, 17(6): 0-362.
[15]	吴月燕;杨祚胜;丁伟红 . 高湿弱光对葡萄叶片光合生理生化指标的影响[J]. , 2005, 17(1): 0-10.

弱光条件下猪舍清洗目标检测

Detection of piggery cleaning target under low-light condition

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 11

参考文献 13

相关文章 15

编辑推荐

Metrics

本文评价