基于增强型轻量深度网络的牧区牲畜高效检测

doi:10.19678/j.issn.1000-3428.0064802

计算机工程 ›› 2023, Vol. 49 ›› Issue (7): 278-287. doi: 10.19678/j.issn.1000-3428.0064802

基于增强型轻量深度网络的牧区牲畜高效检测

齐咏生¹^,², 杜晓旭¹, 朱俊峰¹^,³, 高胜利⁴, 刘利强¹^,²

1. 内蒙古自治区机电控制重点实验室, 呼和浩特 010051
2. 内蒙古工业大学电力学院, 呼和浩特 010051
3. 水利部牧区水利科学研究所, 呼和浩特 010051
4. 内蒙古北方龙源风力发电有限责任公司, 呼和浩特 010050

收稿日期:2022-05-25 出版日期:2023-07-15 发布日期:2023-07-14
作者简介:
齐咏生(1975—)，男，教授、博士，主研方向为目标检测
杜晓旭，硕士研究生
朱俊峰，硕士
高胜利，博士
刘利强，教授、博士
基金资助:
国家自然科学基金(61763037); 内蒙古自治区机电控制重点实验室开放基金(IMMEC2020001); 内蒙古自治区科技计划项目(2021GG164); 内蒙古自治区自然科学基金(2021MS06018); 内蒙古自治区2022年度基本科研业务费项目(JY20220353)

Efficient Livestock Detection in Grazing Areas Based on Enhanced Lightweight Deep Network

Yongsheng QI¹^,², Xiaoxu DU¹, Junfeng ZHU¹^,³, Shengli GAO⁴, Liqiang LIU¹^,²

1. Inner Mongolia Key Laboratory of Electrical and Mechanical Control, Hohhot 010051, China
2. Institute of Electric Power, Inner Mongolia University of Technology, Hohhot 010051, China
3. Pastoral Water Conservancy Science Research Institute of the Ministry of Water Resources, Hohhot 010051, China
4. Inner Mongolia North Longyuan Wind Power Generation Co., Ltd., Hohhot 010050, China

Received:2022-05-25 Online:2023-07-15 Published:2023-07-14

摘要/Abstract

摘要：

实现大数据管理牲畜需要实时监测牲畜，但对牲畜进行实时监测容易受到目标尺寸变化大、光照、环境因素等干扰，因此检测难度大，现有牲畜检测算法存在鲁棒性差等问题。提出一种基于增强型YOLOv4-tiny的目标检测算法(E-YOLOv4-tiny)，采用多尺度特征融合的金字塔网络，兼顾浅层局部细节特征与深层语义信息，解决牧区牲畜尺寸波动问题。通过改进残差结构，减少主干网络参数量，以适应嵌入式平台需求。引入一种新的复合聚类算法设计锚框，在保证可移植性的前提下提高算法精度。针对牧区环境特点，提出一种新的复合多通道注意力机制，改善目标检测网络精度差的问题，增强算法鲁棒性。实验结果表明，E-YOLOv4-tiny算法的平均精度均值(mAP)为0.878 9，帧率为32帧/s，相较于传统YOLOv4-tiny算法，在保持几乎相同的检测速率条件下，mAP提升了9.32%。

关键词: 目标检测, 深度学习, 计算机视觉, YOLOv4-tiny算法, 注意力机制, 特征融合

Abstract:

Realizing big data to manage livestock requires real-time monitoring of livestock, but real-time monitoring of livestock is easily interfered by large changes in target size, lighting, environmental factors, etc., so it is difficult to detection, and existing livestock detection algorithms have the problem of poor robustness.An object detection network called E-YOLOv4-tiny is proposed based on enhanced YOLOv4-tiny, which adopts a pyramid network with multi-scale feature fusion, taking into account shallow local detail features and deep semantic information to solve the problem of livestock size fluctuation in pastoral areas.The number of backbone network parameters is reduced by improving the residual structure to accommodate embedded platform requirements.A new composite clustering algorithm is introduced to design anchor frames to improve the accuracy of the algorithm under the premise of ensuring portability.Finally, according to the characteristics of a pastoral environment, a new Compound Muti-channel Attention(CMA) mechanism is proposed to improve the poor accuracy of the target detection network and enhance the robustness of the algorithm.Experimental results show that the mean Average Precision(mAP) of the E-YOLOv4-tiny algorithm is 0.878 9, and the frame rate is 32 frame/s, and it's mAP is 9.32% higher than that of the traditional YOLOv4-tiny algorithm while maintaining almost the same detection rate.

Key words: target detection, deep learning, computer vision, YOLOv4-tiny algorithm, attention mechanism, feature fusion

齐咏生, 杜晓旭, 朱俊峰, 高胜利, 刘利强. 基于增强型轻量深度网络的牧区牲畜高效检测[J]. 计算机工程, 2023, 49(7): 278-287.

Yongsheng QI, Xiaoxu DU, Junfeng ZHU, Shengli GAO, Liqiang LIU. Efficient Livestock Detection in Grazing Areas Based on Enhanced Lightweight Deep Network[J]. Computer Engineering, 2023, 49(7): 278-287.

https://www.ecice06.com/CN/Y2023/V49/I7/278

图/表 14

图1 E-YOLOv4-tiny网络的结构

Fig.1 Structure of E-YOLOv4-tiny network

图2 算法优化的流程

Fig.2 Procedure of algorithm optimization

图3 通道注意力机制的原理

Fig.3 Principle of squeeze and excitation mechanism

图4 CMA机制的示意图

Fig.4 Schematic diagram of CMA mechanism

图5 残差网络模块的优化

Fig.5 Optimization of residual network module

图6 本文测试平台

Fig.6 Test platform in this paper

表1 消融实验结果

Table 1 Results of ablation experiment

锚框聚类算法			注意力机制		结构优化		性能指标
DBSCAN	K-means++	本文算法	SE	CMA	改进残差网络	多尺度预测	Loss函数均值	mAP	参数量/M
√	×	×	×	×	×	×	0.637	0.7971	23.5
×	√	×	×	×	×	×	0.511	0.8121	23.5
×	×	√	×	×	×	×	0.497	0.8283	23.5
×	×	√	√	×	×	×	0.474	0.8231	25.4
×	×	√	×	√	×	×	0.328	0.8598	27.8
×	×	√	×	√	√	×	0.331	0.8588	27.1
×	×	√	×	√	√	√	0.291	0.8789	29.4

图7 不同算法的Loss函数曲线

Fig.7 Loss function curves of different algorithms

图8 不同算法的PR曲线

Fig.8 PR curves of different algorithms

表2 不同算法的Loss函数均值及mAP值的对比结果

Table 2 Comparison results of average Loss function value and mAP value of different algorithms

算法	Loss函数均值	mAP
YOLOv4	0.189 0	0.910 9
YOLOv4-tiny	0.637 0	0.803 9
E-YOLOv4-tiny	0.290 9	0.878 9
YOLOv5-S	—	0.859 0
YOLOX-S	—	0.861 0

图9 不同算法对牛的目标检测效果对比

Fig.9 Comparison of target detection effects of different algorithms for cows

图10 不同算法对马的目标检测效果对比

Fig.10 Comparison of target detection effects of different algorithms for horses

图11 不同算法对羊的目标检测效果对比

Fig.11 Comparison of target detection effects of different algorithms for sheeps

表3 不同算法的参数对比

Table 3 Comparison of parameters between different algorithms

算法	卷积层数	模型大小/MB	帧率/(帧·s^-1)
YOLOv4	162	256	8
YOLOv4-tiny	38	23.5	39
E-YOLOv4-tiny	61	27.4	32
YOLOv5-S	—	27.5	41
YOLOX-S	—	36.0	33

参考文献 25

1	YI F, JEONG O, MOON I, et al. Deep learning integral imaging for three-dimensional visualization, object detection, and segmentation. Optics and Lasers in Engineering, 2021, 146(3): 106695- 106702.
2	HE K M, GKIOXARI G, DOLLÁR P, et al. Mask R-CNN[C]//Proceedings of IEEE International Conference on Computer Vision. Washington D. C., USA: IEEE Press, 2017: 2980-2988.
3	KIM C J, LEE M J, HWANG K H, et al. End-to-end deep learning-based autonomous driving control for high-speed environment. The Journal of Supercomputing, 2022, 78(2): 1961- 1982. doi: 10.1007/s11227-021-03929-8
4	LORENTE P J N, MILLER L, ROSIQUE F, et al. End-to-end deep neural network architectures for speed and steering wheel angle prediction in autonomous driving. Electronics, 2021, 10(11): 1266- 1272. doi: 10.3390/electronics10111266
5	XUE Y, JU Z. Multiple pedestrian tracking under first-person perspective using deep neural network and social force optimization. Optik, 2021, 240(6): 81- 92.
6	GAO H, WANG W, YANG C, et al. Traffic signal image detection technology based on YOLO. Journal of Physics: Conference Series, 2021, 61(1): 12- 21.
7	NIU H C, HU X B, LI H. Improved YOLOv5 network-based object detection for anti-intrusion of gantry crane[C]//Proceedings of the 2nd International Conference on Control, Robotics and Intelligent System. New York, USA: ACM, 2021: 147-152.
8	REN S Q, HE K M, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[C]//Proceedings of IEEE Transactions on Pattern Analysis and Machine Intelligence. Washington D. C., USA: IEEE Press, 2016: 1137-1149.
9	兰玉彬, 赵德楠, 张彦斐, 等. 生态无人农场模式探索及发展展望. 农业工程学报, 2021, 37(9): 312- 327. URL
	LAN Y B, ZHAO D N, ZHANG Y F, et al. Exploration and development prospect of eco-unmanned farm modes. Transactions of the Chinese Society of Agricultural Engineering, 2021, 37(9): 312- 327. URL
10	王斯佳. 基于智能项圈的畜牧健康诊断算法的研究[D]. 哈尔滨: 哈尔滨工业大学, 2019.
	WANG S J. Research on animal husbandry health diagnosis algorithm based on smart collar[D]. Harbin: Harbin Institute of Technology, 2019. (in Chinese)
11	张磊, 许静, 田在, 等. 牧区智能数羊器系统的设计与实现. 通信电源技术, 2017, 34(4): 165- 166. URL
	ZHANG L, XU J, TIAN Z, et al. Research and implementation of intelligent counting sheep system in pastoral areas. Telecom Power Technology, 2017, 34(4): 165- 166. URL
12	TASSINARI P, BOVO M, BENNI S, et al. A computer vision approach based on deep learning for the detection of dairy cows in free stall barn. Computers and Electronics in Agriculture, 2021, 182(7): 30- 38.
13	WU D, WU Q, YIN X, et al. Lameness detection of dairy cows based on the YOLOv3 deep learning algorithm and a relative step size characteristic vector. Biosystems Engineering, 2020, 189(11): 150- 163.
14	ACHOUR B, BELKADI M, FILALI I, et al. Image analysis for individual identification and feeding behaviour monitoring of dairy cows based on convolutional neural networks. Biosystems Engineering, 2020, 198(7): 31- 49.
15	MA L Z, CHEN Y, ZHANG J L. Vehicle and pedestrian detection based on improved YOLOv4-tiny model. Journal of Physics: Conference Series, 2021, 1920(1): 34- 43.
16	HUANG B Q, LIN H P, HU Z J, et al. An improved YOLOv3-tiny algorithm for vehicle detection in natural scenes. IET Cyber-Systems and Robotics, 2021, 32(3): 256- 264.
17	GUO C, LV X L, ZHANG Y, et al. Improved YOLOv4-tiny network for real-time electronic component detection. Scientific Reports, 2021, 11(1): 44- 56. doi: 10.1038/s41598-020-78528-3
18	LEE G, HONG S, CHO D. Self-supervised feature enhancement networks for small object detection in noisy images. IEEE Signal Processing Letters, 2021, 28(7): 1026- 1030.
19	MA A L, WANG J J, ZHONG Y F, et al. FactSeg: foreground activation-driven small object semantic segmentation in large-scale remote sensing imagery. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60(12): 1- 16.
20	ZHANG Y, LI K. Image super-resolution using very deep residual channel attention networks[EB/OL]. [2022-04-20]. https://arxiv.org/abs/1807.02758.
21	WANG C Y, MARK LIAO H Y, WU Y H, et al. CSPNet: a new backbone that can enhance learning capability of CNN[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Washington D. C., USA: IEEE Press, 2020: 1571-1580.
22	BOCHKOVSKIY A, WANG C Y, LIAO H Y M. YOLOv4: optimal speed and accuracy of object detection[EB/OL]. [2022-04-20]. https://arxiv.org/abs/2004.10934.
23	宋怀波, 江梅, 王云飞, 等. 融合卷积神经网络与视觉注意机制的苹果幼果高效检测方法. 农业工程学报, 2021, 37(9): 297- 303. URL
	SONG H B, JIANG M, WANG Y F, et al. Efficient detection method for young apples based on the fusion of convolutional neural network and visual attention mechanism. Transactions of the Chinese Society of Agricultural Engineering, 2021, 37(9): 297- 303. URL
24	GLENN J. YOLOV5. https://github.com/ultralytics/yolov5.
25	GE Z, LIU S, WANG F, et al. YOLOX: exceeding YOLO series in 2021[EB/OL]. [2022-04-20]. https://arxiv.org/abs/2107.08430.

[1]	王炼红, 林飞鹏, 李潇瑶, 谌桂枝, 周莉. 融入课程知识图谱的KMAKT预测[J]. 计算机工程, 2024, 50(7): 23-31.
[2]	谭巨全, 王然. 特征融合下田径录像3D人体动作DTW捕捉算法[J]. 计算机工程, 2024, 50(7): 71-78.
[3]	钱来, 赵卫伟. 基于对比学习和注意力机制的文本分类方法[J]. 计算机工程, 2024, 50(7): 104-111.
[4]	刘建敏, 林晖, 汪晓丁. 基于图注意力机制的无地图场景轨迹预测方法[J]. 计算机工程, 2024, 50(7): 144-153.
[5]	牛瑞婷, 严天峰, 高锐, 王映植. 低信噪比下基于深度学习TCNN-MobileNet的调制识别[J]. 计算机工程, 2024, 50(7): 204-215.
[6]	屠乃威, 焦猛, 阎馨. 复杂环境下输电线路鸟巢目标图像检测模型[J]. 计算机工程, 2024, 50(7): 216-226.
[7]	张溢文, 蔡满春, 陈咏豪, 朱懿, 姚利峰. 融合空间特征的多尺度深度伪造检测方法[J]. 计算机工程, 2024, 50(7): 240-250.
[8]	肖慈, 徐杨, 张永丹, 冯明文, 黄易仟. 结合注意力和低光增强的夜间语义分割[J]. 计算机工程, 2024, 50(7): 271-281.
[9]	张诗婧, 莫绪涛, 赵行, 董杨林. 基于球面折反射成像和YOLOv7的内螺纹缺陷检测[J]. 计算机工程, 2024, 50(7): 282-292.
[10]	张锡英, 孙守东, 于海浩, 边继龙. 基于空间传播的多视图三维重建[J]. 计算机工程, 2024, 50(7): 293-302.
[11]	王晋涛, 秦昂, 张元, 陈一飞, 王廷凤, 谢承霖, 邹刚. 基于注意力增强与特征融合的中文医学实体识别[J]. 计算机工程, 2024, 50(7): 324-332.
[12]	贵向泉, 刘世清, 李立, 秦庆松, 李唐艳. 基于改进YOLOv8的景区行人检测算法[J]. 计算机工程, 2024, 50(7): 342-351.
[13]	徐明亮, 李芳媛, 马浩然, 何飞. 大规模神经记录的峰电位聚类算法(特邀)[J]. 计算机工程, 2024, 50(6): 1-34.
[14]	魏琢艺, 罗迈, 李文兵, 曾远松, 余伟江, 杨跃东. 基于多源域适应的单细胞智能分类[J]. 计算机工程, 2024, 50(6): 48-55.
[15]	李亚康, 陈刚. 小角中子散射物理模型自动化筛选[J]. 计算机工程, 2024, 50(6): 56-64.

选择文件类型/文献管理软件名称

选择包含的内容

基于增强型轻量深度网络的牧区牲畜高效检测

Efficient Livestock Detection in Grazing Areas Based on Enhanced Lightweight Deep Network

RichHTML

PDF

可视化

被引次数

摘要/Abstract

引用本文

使用本文

图/表 14

参考文献 25

相关文章 15

编辑推荐

Metrics

本文评价

模态框（Modal）标题

选择文件类型/文献管理软件名称

选择包含的内容

基于增强型轻量深度网络的牧区牲畜高效检测

Efficient Livestock Detection in Grazing Areas Based on Enhanced Lightweight Deep Network

RichHTML

PDF

可视化

被引次数

摘要/Abstract

引用本文

使用本文

图/表 14

参考文献 25

相关文章 15

编辑推荐

Metrics

本文评价