On-site monitoring technology of illegal swimming and fishing based on YoloX-ECA

doi:10.11996/JG.j.2095-302X.2023030465

Abstract

Abstract:

Every year in China, a large number of people die from drowning due to illegal swimming and fishing in reservoirs, rivers, and lakes. These bodies of water are often located in remote areas, making it difficult for staff to supervise them 24 hours a day. Existing target detection methods are either too slow or inaccurate, too large to deploy, or incapable of detecting illegal swimming and fishing in real-time. To address these shortcomings, we proposed the YoloX-ECA model with an attention model. By adding the efficient channel attention (ECA) block to the CSPLayer and FPN, we aimed to improve detection performance for swimming and fishing while maintaining detection speed. Experimental results on self-made datasets showed that the YoloX-ECA achieved over 90% AP for the detection of swimming and fishing classes, with a detection speed of 62.29 fps. Compared with YoloX, mAP was increased by 1.21%. Furthermore, YoloX-ECA’s performance also outperformed other target detection algorithms such as Faster-RCNN. The improved YoloX-ECA model achieved the expected design goals and displayed great prospects for application in the field of intelligent supervision of rivers and lakes.

Key words: YoloX, ECA, object detection, attention model, river and lake supervision

CLC Number:

TP391

LUO Wen-yu, FU Ming-yue. On-site monitoring technology of illegal swimming and fishing based on YoloX-ECA[J]. Journal of Graphics, 2023, 44(3): 465-472.

Figures/Tables 13

Fig. 1 YoloX decoupled head compared with coupled head

Fig. 2 SEnet structure

Fig. 3 Cbam Spatial Attention Module network structure

Fig. 4 ECA network structure

Fig. 5 YoloX network structure with attention mechanism

Table 1 Experimental environment

硬件设备	配置
GPU	RTX 2060
显存	6 GB
软件	版本
Pytorch	1.9
CUDA	10.2
Python	3.8

Fig. 6 Training loss curves ((a) YoloX-ECA; (b) YoloX-CBAM; (c) YoloX-SE models)

Fig. 7 Precision curves ((a) Swimming; (b) Fishing; (c) Warning)

Table 2 Precisio, Recall and AP of YoloX-ECA, YoloX-CBAM, YoloX-SE, YoloX

检测对象	模型	Precisio (%)	Recall (%)	AP (%)
Swimming	YoloX-ECA	81.29	93.75	94.17
	YoloX-CBAM	77.21	90.29	91.75
	YoloX	81.23	91.69	93.62
	YoloX-SE	77.37	90.78	92.63
Fishing	YoloX-ECA	87.17	95.96	96.70
	YoloX-CBAM	83.92	93.92	93.19
	YoloX	83.04	94.95	93.25
	YoloX-SE	78.85	91.94	89.33
Warning	YoloX-ECA	77.75	79.09	82.80
	YoloX-CBAM	82.63	80.11	85.63
	YoloX	81.82	74.55	83.18
	YoloX-SE	79.11	78.18	83.98

Fig. 8 F1-score histogram ((a) YoloX-ECA; (b) YoloX-CBAM; (c) YoloX-SE; (d) YoloX models)

Fig. 9 mAP of YoloX-ECA, YoloX-CBAM, YoloX-SE, YoloX and Faster-RCNN

Fig. 10 Example of YoloX-ECA model detection (((a), (b) Swimming predictions; (c) Swimming and Warning predictions; (d), (e), (f) Warning predictions; (g), (h) Fishing predictions))

Fig. 11 Comparison of detection results ((a) YoloX-ECA; (b) Faster-RCNN)

References 22

[1]	赵宇飞, 刘彪, 王毅, 等. 基于数字图像处理的土石坝坝料合格性智能检测方法[J]. 水利学报, 2022, 53(10): 1194-1206.
	ZHAO Y F, LIU B, WANG Y, et al. Intelligent detection method for material qualification of earth-rock dam based on digital image processing[J]. Journal of Hydraulic Engineering, 2022, 53(10): 1194-1206. (in Chinese)
[2]	吴海燕, 李效宁. 图像识别在甘肃智慧水利中的应用[J]. 中国新通信, 2022, 24(8): 70-74.
	WU H Y, LI X N. Application of image recognition in Gansu wisdom water conservancy[J]. China New Telecommunications, 2022, 24(8): 70-74. (in Chinese)
[3]	赵薛强, 凌峻. 无人机自动巡检智慧监控系统研究与应用[J]. 人民长江, 2022, 53(6): 235-241.
	ZHAO X Q, LING J. Development and application of intelligent monitoring system with UAV automatic inspection[J]. Yangtze River, 2022, 53(6): 235-241. (in Chinese)
[4]	LI J, WANG J Z, LIU W X. Moving target detection and tracking algorithm based on context information[J]. IEEE Access, 2019, 7: 70966-70974. DOI URL
[5]	GU B, HU H, REN Y, et al. Moving target detection and tracking in complex background[J]. International Journal of Smart Home, 2015, 9(9): 95-102. DOI URL
[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. New York: IEEE Press, 2016: 779-788.
[7]	王银, 王飞翔, 孙前来. 多尺度特征融合车辆检测方法[J]. 系统仿真学报, 2022, 34(6): 1219-1229. DOI
	WANG Y, WANG F X, SUN Q L. Vehicle detection method based on multi scale feature fusion[J]. Journal of System Simulation, 2022, 34(6): 1219-1229. (in Chinese) DOI
[8]	张明臻. 基于Dense-YOLO网络的井下行人检测模型[J]. 工矿自动化, 2022, 48(3): 86-90.
	ZHANG M Z. Underground pedestrian detection model based on Dense-YOLO network[J]. Industry and Mine Automation, 2022, 48(3): 86-90. (in Chinese)
[9]	刘力, 苟军年. 基于Yolov4的铁道侵限障碍物检测方法研究[J]. 铁道科学与工程学报, 2022, 19(2): 528-536.
	LIU L, GOU J N. Research on detection method of railway intrusion obstacles based on Yolov4[J]. Journal of Railway Science and Engineering, 2022, 19(2): 528-536. (in Chinese)
[10]	李坤, 樊宇. 基于改进卷积神经网络的船舶图像识别研究[J]. 舰船科学技术, 2021, 43(12): 187-189.
	LI K, FAN Y. Research on ship image recognition based on improved convolution neural network[J]. Ship Science and Technology, 2021, 43(12): 187-189. (in Chinese)
[11]	GE Z, LIU S T, WANG F, et al. YOLOX: exceeding YOLO series in 2021[EB/OL]. (2021-08-06) [2022-07-01]. https://arxiv.org/abs/2107.08430.
[12]	REDMON J, FARHADI A. YOLOv3: an incremental improvement[EB/OL]. (2018-04-08) [2022-07-01]. https://arxiv.org/abs/1804.02767.
[13]	BOCHKOVSKIY A, WANG C Y, LIAO H Y M. YOLOv4: optimal speed and accuracy of object detection[EB/OL]. (2020-04-23) [2022-10-01]. https://arxiv.org/abs/2004.10934.
[14]	CHEN Q, WANG Y M, YANG T, et al. You only look one-level feature[EB/OL]. (2021-05-17) [2022-07-01]. https://arxiv.org/abs/2103.09460.
[15]	GAO W W, SHAN M T, SONG N, et al. Detection of microaneurysms in fundus images based on improved YOLOv4 with SENet embedded[J]. Journal of Biomedical Engineering, 2022, 39(4): 713-720.
[16]	赵杰, 李絮, 申通. 基于SENet注意力机制和深度残差网络的腹部动脉分割[J]. 科学技术与工程, 2022, 22(22): 9529-9536.
	ZHAO J, LI X, SHEN T. Abdominal artery segmentation based on SENet attention mechanism and deep residual network[J]. Science Technology and Engineering, 2022, 22(22): 9529-9536. (in Chinese)
[17]	刘学平, 李玙乾, 刘励, 等. 嵌入SENet结构的改进YOLOV3目标识别算法[J]. 计算机工程, 2019, 45(11): 243-248. DOI
	LIU X, LI Y, LIU L, et al. Improved YOLOV3 target recognition algorithm with embedded SENet structure[J]. Computer Engineering, 2019, 45(11): 243-248. (in Chinese) DOI
[18]	李克文, 李新宇. 基于SENet改进的Faster R-CNN行人检测模型[J]. 计算机系统应用, 2020, 29(4): 266-271.
	LI K W, LI X Y. Pedestrian detection model based on improved faster R-CNN with SENet[J]. Computer Systems & Applications, 2020, 29(4): 266-271. (in Chinese)
[19]	WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module[EB/OL]. (2018-07-18) [2022-07-01]. https://arxiv.org/abs/1807.06521.
[20]	付国栋, 黄进, 杨涛, 等. 改进CBAM的轻量级注意力模型[J]. 计算机工程与应用, 2021, 57(20): 150-156. DOI
	FU G, HUANG J, YANG T, et al. Improved lightweight attention model based on CBAM[J]. Computer Engineering and Applications, 2021, 57(20): 150-156. (in Chinese) DOI
[21]	WANG Q L, WU B G, ZHU P F, et al. ECA-net: efficient channel attention for deep convolutional neural networks[EB/OL]. (2020-03-24) [2022-07-01]. https://arxiv.org/abs/1910.03151.
[22]	HE K M, ZHANG X Y, REN S Q, et al. Delving deep into rectifiers: surpassing human-level performance on ImageNet classification[EB/OL]. (2015-02-06) [2022-07-01]. https://arxiv.org/abs/1502.01852.