复杂战场环境下改进YOLOv5军事目标识别算法研究

doi:10.12382/bgxb.2022.0736

摘要/Abstract

摘要：

复杂战场环境下军事目标识别技术是提升战场情报获取能力的基础和关键。针对当前军事目标识别技术在复杂战场环境下漏检误检率高、实时性差等问题,提出一种基于改进YOLOv5模型的PB-YOLO军事目标识别算法。将改进的目标识别算法对于陆战场军事单元的识别锚框进行重新聚类,以提升模型对于目标大小适应度,加速模型收敛;采用通道-空间并行注意力机制,增加模型对复杂战场环境下目标特征信息与位置信息关注度;在特征融合网络部分使用BiFPN以提升模型对于特征的融合能力与速度;采用Alpha_IoU损失函数加速模型收敛,解决当真实框与预测框重合时IoU计算退化问题。实验结果表明,在自建军事目标数据集下,改进算法与主流目标识别算法相比,在保证模型空间复杂度的同时,mAP值达到了90.17%。消融实验对比结果表明,改进后网络较原模型精度提升11.57%,具有较好的识别性能,能够为战场情报获取提供有效的技术支撑。

关键词: 军事目标识别, 通道-空间并行注意力机制, 特征融合, 损失函数

Abstract:

Military target recognition technology used in complex battlefield environment is the basis and key to improve the battlefield intelligence acquisition capability. A PB-YOLO military target recognition algorithm based on the improved YOLOv5 model is proposed to solve the problems of high missed and false detection rates and poor real-time performance of current military target recognition technology in complex battlefield environments. The improved target recognition algorithm is re-clustered for the identification anchor boxes of military units in the land battlefield to improve the model’s fitness for target size and accelerate the convergence of model, and the channel-spatial parallel attention mechanism is used to increase the model’s attention to the feature information and location information of the targets in complex battlefield environments. BiFPN is used in the feature fusion network part to improve the fusion ability and speed of the model for features, and the Alpha_IoU loss function is used to accelerate the convergence of model, and solve the problem of IoU calculation degradation when the real frame and the predicted frame overlap. The experimental results show that,compared with the mainstream target recognition algorithm, the mAP value obtained by the improved algorithm reaches 90.17% while ensuring the model space complexity under the self-built military target data set., Through the comparison of ablation experiments, the results show that, compared with the original model, the accuracy of the improved network is improved by 11.57%, and it has better recognition performance, which can provide effective technical support for battlefield intelligence acquisition.

Key words: military target recognition, channel-spatial parallel attention mechanism, feature fusion, loss function

中图分类号:

TP391.4

宋晓茹, 刘康, 高嵩, 陈超波, 阎坤. 复杂战场环境下改进YOLOv5军事目标识别算法研究[J]. 兵工学报, 2024, 45(3): 934-947.

SONG Xiaoru, LIU Kang, GAO Song, CHEN Chaobo, YAN Kun. Research on Improved YOLOv5-based Military Target Recognition Algorithm Used in Complex Battlefield Environment[J]. Acta Armamentarii, 2024, 45(3): 934-947.

图/表 19

图1 YOLOv5s网络结构图

Fig.1 YOLOv5s network structure diagram

图2 PB-YOLO网络结构图

Fig.2 PB-YOLO network structure diagram

图3 通道-空间并行注意力机制引入位置图

Fig.3 Channel-Spatial parallel attention mechanism introduces location map

图4 通道-空间注意力机制结构图

Fig.4 Structure diagram of channel-spatial attention mechanism

图5 第3层~第7层采用不同特征融合网络示意图

Fig.5 Schematic diagram of different feature fusion networks from Layer 3 to Layer 7

表1 初始参数设置

Table 1 Initial parameter setting table

参数	数值
lr	0.01
momentum	0.937
weight_decay	0.0005
batch_size	16
image_size	640×640
epoch	300

图6 P450-NX科研无人机

Fig.6 P450-NX UAV

图7 部分数据集样图

Fig.7 Sample drawings of some datasets

表2 目标数量及尺寸统计

Table 2 Statistical table of target quantity and size

数量及尺寸	坦克	士兵	装甲车	自行火炮
数量/个	1652	2497	1086	1087
尺寸/m	11×3.4×2		5.5×2.5×2.3	8×3×2.1

表3 重聚类后先验框宽高值

Table 3 Prior box width and height values after reclustering

序号	先验框宽高值
1	[12.734 28.496]
2	[18.768 39.405]
3	[52.218 42.258]
4	[35.064 83.506]
5	[102.03 61.18]
6	[153.82 115.14]
7	[232.96 306.19]
8	[409.64 177.11]
9	[526.23 396.53]

图8 主流算法精度对比图

Fig.8 Precision comparison diagram of mainstream algorithms

图9 主流算法识别效果对比图

Fig.9 Comparison of recognition effects of mainstream algorithms

表4 主流算法识别结果对比

Table 4 Comparison of recognition results of mainstream algorithms

主流算法	模型体积/MB	FPS	mAP/%
Faster R-CNN	159.7	6.5	86.7
SSD	23.1	38	83.4
ShuffleNet	3.0	28	73.3
MobileNet	8.0	34	74.8
YOLOv3	235.6	20	82.1
YOLOv4	244.8	48	83.1
YOLOv5m	118.3	44.3	81.5
YOLOv5l	226.7	43.2	83.4
YOLOv5x	401.0	40.9	84.0
YOLOv5s+SE	60.0	42	81.1
YOLOv5s+CBAM	63.0	37	84.0
PB-YOLO(无BiFPN)	63.3	51	85.0
PB-YOLO	64.0	57	90.17

表5 PB-YOLO消融实验对比

Table 5 PB-YOLO ablation experiment comparison

模型	识别算法及改进	Parameters/M	GFLOPs	Training time/h	mAP/%
A	YOLOv5	7.021	15.8	4.12	78.6
B	A+通道注意力机制	6.932	15.5	4.5	80.4
C	A+空间注意力机制	6.879	15.7	4.37	81.2
D	A+并行注意力机制	6.713	15.3	5.12	84.3
E	A+Alpha-IoU	7.021	15.8	3.495	79.3
F	B+Alpha-IoU	6.932	15.5	4.39	80.8
G	C+Alpha-IoU	6.879	15.7	4.22	81.5
H	D+Alpha-IoU	6.713	15.3	4.75	85.0
I	A+BiFPN	7.169	16.5	2.126	83.6
J	D+BiFPN	6.862	15.9	3.304	89.47
K	E+BiFPN	7.169	16.5	2.061	85.5
L	H+BiFPN	6.862	15.9	3.261	90.17

图10 消融实验对比图

Fig.10 Comparison of ablation experiments

图11 Alpha-IoU精度提升对比图

Fig.11 Alpha-IoU accuracy improvement comparison chart

图12 单个目标不同算法识别对比图

Fig.12 Comparison of identification of different algorithms for a single target

图13 多目标识别对比图

Fig.13 Multi-target recognition comparison chart

图14 陆战场军事目标实时识别图

Fig.14 Real-time identification map of military targets in land battlefield

参考文献 28

[1]	CHEN C, HUANG J, PAN C Y, et al. Military image scene recognition based on CNN and semantic information[C]//Proceedings of the 2018 3rd International Conference on Mechanical, Control and Computer Engineering.Huhhot,China:IEEE, 2018: 573-577.
[2]	于博文, 吕明. 改进的YOLOv3算法及其在军事目标检测中的应用[J]. 兵工学报, 2022, 43(2):345-354. doi: 10.3969/j.issn.1000-1093.2022.02.012
	YU B W, LÜ M. Improved YOLOv3 algorithm and its application in military target detection[J]. Acta Armamentarii, 2022, 43(2):345-354. (in Chinese) doi: 10.3969/j.issn.1000-1093.2022.02.012
[3]	刘俊, 孟伟秀, 余杰, 等. 面向军事目标识别的DRFCN深度网络设计及实现[J]. 光电工程, 2019, 46(4):21-30.
	LIU J, MENG W X, YU J, et al. Design and implementation of DRFCN deep network for military target recognition[J]. Optoelectronic Engineering, 2019, 46(4): 21-30. (in Chinese)
[4]	GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Piscataway, NJ,US:IEEE, 2014:580-587.
[5]	WANG Z Z, XIE K, ZHANG X Y, et al. Small-object detection based on YOLO and dense block via image super-resolution[J]. IEEE Access, 2021, 9: 56416-56429. doi: 10.1109/ACCESS.2021.3072211 URL
[6]	GIRSHICK R. Fast R-CNN[C]//Proceedings of International Conference on Computer Vision. Piscataway, NJ, US: IEEE, 2015:1440-1448.
[7]	OTHMANI M. A vehicle detection and tracking method for traffic video based on faster R-CNN[J]. Multimedia Tools and Applications, 2022, 81(20):28347-28365. doi: 10.1007/s11042-022-12715-4
[8]	SI J X, LIN J J, JIANG F, et al. Hand-raising gesture detection in real classrooms using improved R-FCN[J]. Neurocomputing, 2019, 359:69-76. doi: 10.1016/j.neucom.2019.05.031
[9]	ZHANG C, ZHOU J W, WANG H W, et al. Multi-species individual tree segmentation and identification based on improved mask R-CNN and UAV imagery in mixed forests[J]. Remote Sensing, 2022, 14(4):874-874. doi: 10.3390/rs14040874 URL
[10]	REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]//Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, NE,US:IEEE, 2016:779-788.
[11]	孙剑明, 韩生权, 沈子成, 等. 基于双卷积链的双目人体姿态距离定位识别[J/OL]. 兵工学报:1-8[2022-08-23]. DOI: 10.12382/bgxb.2021.0629.
	SUN J M, HAN S Q, SHEN Z C, et al. Binocular human body pose distance localization and recognition based on double convolution chain[J/OL]. Acta Armamentarii:1-8[2022-08-23].DOI:10.12382/bgxb.2021.0629. (in Chinese)
[12]	XU M J, WANG X H, ZHANG S, et al. Detection algorithm of aerial vehicle target based on improved YOLOv3[J]. Journal of Physics:Conference Series, 2022, 2284(1): 012-022.
[13]	KONG H, CHEN Z, YUE W J, et al. Improved YOLOv4 for pedestrian detection and counting in UAV images[J]. Computational Intelligence and Neuroscience, 2022, 2022:6106853-6106853.
[14]	LIU G, CAO Z X, LIU S, et al. An improved SSD method for infrared target detection based on convolutional neural network[J]. Journal of Computational Methods in Sciences and Engineering, 2022, 22(4):1393-1408. doi: 10.3233/JCM-226112 URL
[15]	LIN T Y, GOYAL P, GRISHICK R, et al. Focal loss for dense object detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(2):318-327. doi: 10.1109/TPAMI.34 URL
[16]	潘浩. 基于深度学习的军事目标识别[D]. 杭州: 杭州电子科技大学, 2018.
	PAN H. Military target recognition based on deep learning[D]. Hangzhou: Hangzhou Dianzi University, 2018. (in Chinese)
[17]	ZENG G Z, SONG R, HU X, et al. pplying convolutional neural network for military object detection on embedded platform[J]. Computer Engineering and Technology, 2018, 22(1): 131-141.
[18]	陈龙, 张峰, 蒋升. 小样本条件下基于深度森林学习模型的典型军事目标识别方法[J]. 中国电子科学研究院学报, 2019, 14(3):232-237.
	CHEN L, ZHANG F, JIANG S. Typical military target recognition method based on deep forest learning model under small sample conditions[J]. Journal of Chinese Academy of Electronic Sciences, 2019, 14(3):232-237. (in Chinese)
[19]	林洋, 董宝良, 刘泽平. 一种基于CGAN和GcForest的军事目标识别方法[J]. 信息技术, 2020, 44(3):134-138.
	LIN Y, DONG B L, LIU Z P. A military target recognition method based on CGAN and GcForest[J]. Information Technology, 2020, 44(3): 134-138. (in Chinese)
[20]	杨朝红, 王伟男. 基于优化SSD300的小尺度典型军事目标识别方法研究[J]. 电脑与信息技术, 2020, 28(4):19-22.
	YANG C H, WANG W N. Research on small-scale typical military target recognition method based on optimized SSD300[J]. Computer and Information Technology, 2020, 28(4): 19-22. (in Chinese)
[21]	李鑫, 王晟全, 李昂. 基于非监督网络的军事目标识别算法的研究[J]. 电光与控制, 2021, 28(10): 36-39.
	LI X, WANG S Q, LI A. Research on military target recognition algorithm based on unsupervised network[J]. Electro Optic and Control, 2021, 28(10):36-39. (in Chinese)
[22]	PARK J, WOO S, LEE J Y, et al. Bam: Bottleneck attention module:arXiv:1807.0 6514[R]. Ithaca,NY,US: Cornell University, 2018:1807.0 6514.
[23]	TAN M X, PANG R M, LE Q V. Efficientdet: Scalable and efficient object detection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle, WA,US:IEEE, 2020: 10781-10790.
[24]	HE J B, ERFANI S, MA X J, et al. alpha-IoU: a family of power intersection over union losses for bounding box regression[J]. Advances in Neural Information Processing Systems, 2021, 34: 20230-20242.
[25]	LÜ H H, YAN H B, LIU K Y, et al. YOLOv5-AC: attention mechan-ism-based lightweight YOL-Ov5 for trck pedestrian detection[J]. Sensors, 2022, 22(15):5903. doi: 10.3390/s22155903 URL
[26]	HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, NV,US:IEEE, 2016: 770-778.
[27]	IOFFE S, SZEGEDY C. Batch normalization:accel-erating deep network training by reduc-ing internal covariateshift[C]//Proceedings of International Conference on Machine Learning. Miami, FL,US:ICMLA, 2015: 448-456.
[28]	任江涛, 李定主, 屠惠琳. 基于训练网络的目标检测方法及应用[J]. 火力与指挥控制, 2020, 45(4):173-177.
	REN J T, LI D Z, TU H L. Target detection method and application based on training network[J]. Firepower and Command and Control, 2020, 45(4):173-177. (in Chinese)