图学学报 ›› 2024, Vol. 45 ›› Issue (6): 1313-1327.DOI: 10.11996/JG.j.2095-302X.2024061313
收稿日期:
2024-07-23
接受日期:
2024-09-26
出版日期:
2024-12-31
发布日期:
2024-12-24
通讯作者:
侯立群(1972-),男,副教授,博士。主要研究方向为图像处理、无线传感器网络和设备故障诊断。E-mail:houliqun@ncepu.edu.cn第一作者:
李盛涛(1998-),男,硕士研究生。主要研究方向为图像处理和深度学习。E-mail:lst18315889026@163.com
基金资助:
LI Shengtao(), HOU Liqun(
), DONG Yasong
Received:
2024-07-23
Accepted:
2024-09-26
Published:
2024-12-31
Online:
2024-12-24
Contact:
HOU Liqun (1972-), associate professor, Ph.D. His main research interests cover image processing, wireless sensor networks, and device fault diagnosis. E-mail:houliqun@ncepu.edu.cnFirst author:
LI Shengtao (1998-), master student. His main research interests cover image processing and deep learning. E-mail:lst18315889026@163.com
Supported by:
摘要:
针对现有方法中表盘关键信息提取过程繁琐、读数误差较大和相机抖动导致的运动模糊问题,提出了一种基于R-YOLOv7和MIMO-CTFNet的指针式仪表自动读数方法。首先,构建兼顾精度和轻量化的R-YOLOv7算法实现指针式仪表表盘和表盘关键信息检测;然后,设计了MIMO-CTFNet算法以实现运动模糊仪表图像的复原;最后,利用提取的表盘关键信息进行基于小刻度线的角度法读数。实验结果表明改进后的R-YOLOv7在表盘关键信息检测数据集上所需的参数量、FLOPs、ADT和mAP50:95分别为12 M个、60.30 G次、17.04 ms和86.5%;改进后的MIMO-CTFNet算法在采集的运动模糊数据集上的PSNR和SSIM分别达到33.05 dB和0.935 3;该读数方法的读数最大引用误差为0.35%,需要运动模糊处理和无需运动模糊处理的图像读数时间分别为0.561 s和0.128 s,从而验证了该方法的有效性。
中图分类号:
李盛涛, 侯立群, 董亚松. 基于R-YOLOv7和MIMO-CTFNet的指针式仪表自动读数方法[J]. 图学学报, 2024, 45(6): 1313-1327.
LI Shengtao, HOU Liqun, DONG Yasong. Automatic reading of pointer meters based on R-YOLOv7 and MIMO-CTFNet[J]. Journal of Graphics, 2024, 45(6): 1313-1327.
图9 LGFEM模块的子模块结构图((a) AU模块结构;(b) ARB模块结构;(c) MDTA模块结构;(d) GDFN模块结构)
Fig. 9 Submodule structure diagram of LGFEM module ((a) AU module structure; (b) ARB module structure; (c) MDTA module structure; (d) GDFN module structure)
图10 运动模糊仪表图像复原((a)运动模糊图像;(b)运动模糊图像修复结果)
Fig. 10 Motion blur instrument image restoration ((a) Motion blur image; (b) Motion blur image restoration results)
图12 刻度值的空间聚合((a)数字和小数点检测结果;(b)数字框合并及刻度值计算结果)
Fig. 12 Spatial aggregation of scale values ((a) Numbers and decimal points detection results; (b) Results of combining number boxes and calculating scale values)
图14 刻度线增补或删除示意图((a)缺失或多出部分刻度线中心点示意图;(b)增补或删除刻度线中心点结果)
Fig. 14 Schematic diagram for adding or deleting scale lines ((a) Schematic diagram of the center point of the missing or extra part of the scale line; (b) Adding or deleting scale center point results)
参数 | 配置 |
---|---|
操作系统 | Ubuntu20.04 |
CPU型号 | Intel(R) Xeon(R) Platinum 8255C |
显卡(GPU)型号 | NVIDIA GeForce RTX2080Ti |
显卡内存 | 11 G |
深度学习框架 | Pytorch 1.10.0 |
编程语言 | Python 3.8 |
表1 实验环境参数
Table 1 Experimental environment parameters
参数 | 配置 |
---|---|
操作系统 | Ubuntu20.04 |
CPU型号 | Intel(R) Xeon(R) Platinum 8255C |
显卡(GPU)型号 | NVIDIA GeForce RTX2080Ti |
显卡内存 | 11 G |
深度学习框架 | Pytorch 1.10.0 |
编程语言 | Python 3.8 |
算法 | 参数量/ M | FLOPs/ G | ADT/ ms | PSNR/ dB | SSIM |
---|---|---|---|---|---|
Baseline | 6.8 | 63.75 | 30.35 | 30.63 | 0.924 7 |
Baseline+A | 12.3 | 104.18 | 126.45 | 31.81 | 0.931 5 |
Baseline+B | 3.8 | 36.42 | 106.87 | 32.49 | 0.930 4 |
Baseline+A+B | 9.3 | 76.85 | 251.78 | 33.05 | 0.935 3 |
表2 不同改进策略的能对比
Table 2 Comparison of the performance of different improvement strategies
算法 | 参数量/ M | FLOPs/ G | ADT/ ms | PSNR/ dB | SSIM |
---|---|---|---|---|---|
Baseline | 6.8 | 63.75 | 30.35 | 30.63 | 0.924 7 |
Baseline+A | 12.3 | 104.18 | 126.45 | 31.81 | 0.931 5 |
Baseline+B | 3.8 | 36.42 | 106.87 | 32.49 | 0.930 4 |
Baseline+A+B | 9.3 | 76.85 | 251.78 | 33.05 | 0.935 3 |
算法 | 参数量/ M | FLOPs/ G | ADT/ ms | PSNR/ dB | SSIM |
---|---|---|---|---|---|
MIMO-UNet | 6.80 | 63.75 | 30.35 | 30.63 | 0.924 7 |
MIMO-UNet+ | 16.10 | 150.72 | 48.82 | 30.95 | 0.930 8 |
SFNet | 13.27 | 125.43 | 252.18 | 31.09 | 0.934 7 |
Ours | 9.30 | 76.85 | 251.78 | 33.05 | 0.935 3 |
表3 不同算法的性能对比
Table 3 Comparison of the performance of different algorithms
算法 | 参数量/ M | FLOPs/ G | ADT/ ms | PSNR/ dB | SSIM |
---|---|---|---|---|---|
MIMO-UNet | 6.80 | 63.75 | 30.35 | 30.63 | 0.924 7 |
MIMO-UNet+ | 16.10 | 150.72 | 48.82 | 30.95 | 0.930 8 |
SFNet | 13.27 | 125.43 | 252.18 | 31.09 | 0.934 7 |
Ours | 9.30 | 76.85 | 251.78 | 33.05 | 0.935 3 |
图19 各个算法图像复原效果对比((a)模糊图片;(b) MIMO-UNet;(c) MIMO-UNet+;(d) SFNet;(e)本文方法;(f)清晰图片)
Fig. 19 Comparison of image restoration results of various algorithms ((a) Blurry picture; (b) MIMO-UNet; (c) MIMO-UNet+; (d) SFNet; (e) Ours; (f) Clear picture)
算法 | 参数量/M | FLOPs/G | ADT/ms |
---|---|---|---|
Model 1 | 37.28 | 105.37 | 18.00 |
Model 2 | 38.22 | 107.80 | 18.18 |
Model 3 | 38.38 | 109.00 | 18.63 |
Model 4 | 12.00 | 60.30 | 17.04 |
表4 不同改进策略参数量、运算量和平均推理时间对比
Table4 Comparison of number of parameters, operations and average reasoning time for different improved strategies
算法 | 参数量/M | FLOPs/G | ADT/ms |
---|---|---|---|
Model 1 | 37.28 | 105.37 | 18.00 |
Model 2 | 38.22 | 107.80 | 18.18 |
Model 3 | 38.38 | 109.00 | 18.63 |
Model 4 | 12.00 | 60.30 | 17.04 |
目标类别 | mAP50:95 | |||
---|---|---|---|---|
Model 1 | Model 2 | Model 3 | Model 4 | |
scale | 62.9 | 80.7 | 83.0 | 82.4 |
pointer | 24.6 | 91.0 | 93.2 | 91.9 |
R_pointer | 47.3 | 90.3 | 93.7 | 90.6 |
0 | 71.7 | 85.3 | 87.3 | 86.5 |
1 | 71.3 | 81.6 | 83.1 | 82.8 |
2 | 79.6 | 86.6 | 88.4 | 88.6 |
3 | 81.5 | 87.7 | 88.9 | 89.8 |
4 | 85.4 | 88.1 | 89.3 | 88.5 |
5 | 81.4 | 86.9 | 88.3 | 87.7 |
6 | 84.6 | 88.2 | 89.6 | 89.4 |
7 | 84.4 | 89.4 | 89.8 | 94.8 |
8 | 86.2 | 90.5 | 91.3 | 91.2 |
9 | 86.3 | 89.6 | 91.3 | 94.2 |
point | 48.4 | 50.3 | 54.6 | 53.0 |
centre | 82.7 | 85.1 | 86.4 | 86.1 |
all | 71.9 | 84.8 | 86.6 | 86.5 |
表5 不同改进策略的检测精度对比/%
Table 5 Comparison of detection accuracy with different improvement strategies/%
目标类别 | mAP50:95 | |||
---|---|---|---|---|
Model 1 | Model 2 | Model 3 | Model 4 | |
scale | 62.9 | 80.7 | 83.0 | 82.4 |
pointer | 24.6 | 91.0 | 93.2 | 91.9 |
R_pointer | 47.3 | 90.3 | 93.7 | 90.6 |
0 | 71.7 | 85.3 | 87.3 | 86.5 |
1 | 71.3 | 81.6 | 83.1 | 82.8 |
2 | 79.6 | 86.6 | 88.4 | 88.6 |
3 | 81.5 | 87.7 | 88.9 | 89.8 |
4 | 85.4 | 88.1 | 89.3 | 88.5 |
5 | 81.4 | 86.9 | 88.3 | 87.7 |
6 | 84.6 | 88.2 | 89.6 | 89.4 |
7 | 84.4 | 89.4 | 89.8 | 94.8 |
8 | 86.2 | 90.5 | 91.3 | 91.2 |
9 | 86.3 | 89.6 | 91.3 | 94.2 |
point | 48.4 | 50.3 | 54.6 | 53.0 |
centre | 82.7 | 85.1 | 86.4 | 86.1 |
all | 71.9 | 84.8 | 86.6 | 86.5 |
算法 | 参数量/ M | FLOPs/ G | ADT/ ms | mAP50:95/ % |
---|---|---|---|---|
R-YOLOv5l | 47.15 | 111.0 | 19.70 | 84.8 |
R-YOLOv5s | 7.54 | 17.4 | 12.25 | 79.2 |
R-YOLOv7-tiny | 6.53 | 14.7 | 11.66 | 76.6 |
Ours | 12.00 | 60.3 | 17.04 | 86.5 |
表6 不同算法的性能对比
Table 6 Comparison of the performance of different algorithms
算法 | 参数量/ M | FLOPs/ G | ADT/ ms | mAP50:95/ % |
---|---|---|---|---|
R-YOLOv5l | 47.15 | 111.0 | 19.70 | 84.8 |
R-YOLOv5s | 7.54 | 17.4 | 12.25 | 79.2 |
R-YOLOv7-tiny | 6.53 | 14.7 | 11.66 | 76.6 |
Ours | 12.00 | 60.3 | 17.04 | 86.5 |
图21 理想环境及有干扰环境下部分指针式仪表自动读数识别过程
Fig. 21 The automatic reading recognition process of partial pointer instruments in both ideal and interference environments
干扰类型 | 序号 | 人工读数/ MPa | 本文方法 读数/MPa | 引用误差/ % |
---|---|---|---|---|
理想环境 | 1 | 0.38 | 0.384 | 0.16 |
2 | 1.72 | 1.722 | 0.08 | |
光线过亮 | 1 | 0.93 | 0.932 | 0.08 |
2 | 1.38 | 1.375 | -0.20 | |
光线过暗 | 1 | 0.38 | 0.383 | 0.12 |
2 | 1.29 | 1.294 | 0.16 | |
光线不均 | 1 | 1.08 | 1.077 | -0.12 |
2 | 2.27 | 2.272 | 0.08 | |
有光斑 | 1 | 1.42 | 1.417 | -0.12 |
2 | 2.17 | 2.175 | 0.20 | |
椒盐噪声 | 1 | 1.38 | 1.377 | -0.12 |
2 | 1.72 | 1.720 | 0.00 | |
倾斜 | 1 | 0.68 | 0.685 | 0.20 |
2 | 2.30 | 2.295 | -0.20 | |
有划痕 | 1 | 1.23 | 1.232 | 0.08 |
2 | 1.83 | 1.832 | 0.08 | |
反光 | 1 | 0.77 | 0.774 | 0.16 |
2 | 1.78 | 1.778 | -0.08 | |
油污 | 1 | 0.26 | 0.262 | 0.08 |
2 | 2.19 | 2.187 | -0.12 |
表7 理想环境及有干扰环境下部分指针式仪表自动读数结果
Table 7 The automatic reading recognition results of partial pointer instruments in both ideal and interference environments
干扰类型 | 序号 | 人工读数/ MPa | 本文方法 读数/MPa | 引用误差/ % |
---|---|---|---|---|
理想环境 | 1 | 0.38 | 0.384 | 0.16 |
2 | 1.72 | 1.722 | 0.08 | |
光线过亮 | 1 | 0.93 | 0.932 | 0.08 |
2 | 1.38 | 1.375 | -0.20 | |
光线过暗 | 1 | 0.38 | 0.383 | 0.12 |
2 | 1.29 | 1.294 | 0.16 | |
光线不均 | 1 | 1.08 | 1.077 | -0.12 |
2 | 2.27 | 2.272 | 0.08 | |
有光斑 | 1 | 1.42 | 1.417 | -0.12 |
2 | 2.17 | 2.175 | 0.20 | |
椒盐噪声 | 1 | 1.38 | 1.377 | -0.12 |
2 | 1.72 | 1.720 | 0.00 | |
倾斜 | 1 | 0.68 | 0.685 | 0.20 |
2 | 2.30 | 2.295 | -0.20 | |
有划痕 | 1 | 1.23 | 1.232 | 0.08 |
2 | 1.83 | 1.832 | 0.08 | |
反光 | 1 | 0.77 | 0.774 | 0.16 |
2 | 1.78 | 1.778 | -0.08 | |
油污 | 1 | 0.26 | 0.262 | 0.08 |
2 | 2.19 | 2.187 | -0.12 |
序号 | 人工读数/ MPa | 本文方法 读数/MPa | 引用误差/ % |
---|---|---|---|
1 | 0.28 | 0.265 | -0.25 |
2 | 2.56 | 2.578 | 0.30 |
3 | 5.36 | 5.351 | -0.15 |
表8 部分单指针式仪表自动读数结果
Table 8 Automatic reading results for some single-pointer meters
序号 | 人工读数/ MPa | 本文方法 读数/MPa | 引用误差/ % |
---|---|---|---|
1 | 0.28 | 0.265 | -0.25 |
2 | 2.56 | 2.578 | 0.30 |
3 | 5.36 | 5.351 | -0.15 |
序号 | 预警值 | 实际值 | ||||
---|---|---|---|---|---|---|
人工 读数/ MPa | 本文 方法 读数/ MPa | 引用 误差/ % | 人工 读数/ MPa | 本文 方法 读数/ MPa | 引用 误差/ % | |
1 | 2.28 | 2.291 | 0.18 | 1.28 | 1.275 | -0.08 |
2 | 5.64 | 5.657 | 0.28 | 3.48 | 3.499 | 0.31 |
3 | 5.84 | 5.819 | -0.35 | 5.36 | 5.355 | -0.08 |
表9 部分双指针式仪表自动读数结果
Table 9 Automatic reading results for some dual-pointer meters
序号 | 预警值 | 实际值 | ||||
---|---|---|---|---|---|---|
人工 读数/ MPa | 本文 方法 读数/ MPa | 引用 误差/ % | 人工 读数/ MPa | 本文 方法 读数/ MPa | 引用 误差/ % | |
1 | 2.28 | 2.291 | 0.18 | 1.28 | 1.275 | -0.08 |
2 | 5.64 | 5.657 | 0.28 | 3.48 | 3.499 | 0.31 |
3 | 5.84 | 5.819 | -0.35 | 5.36 | 5.355 | -0.08 |
组号 | 序号 | 人工读数/ MPa | 本文方法 读数/MPa | 引用误差/ % |
---|---|---|---|---|
指针表1 | 1 | 35.2 | 35.272 | 0.12 |
2 | 44.8 | 44.642 | -0.26 | |
指针表2 | 1 | 40.2 | 40.069 | -0.22 |
2 | 46.8 | 46.793 | -0.01 |
表10 工业场景下部分指针式仪表自动读数结果
Table 10 Automatic reading results of some pointer meters in industrial scenarios
组号 | 序号 | 人工读数/ MPa | 本文方法 读数/MPa | 引用误差/ % |
---|---|---|---|---|
指针表1 | 1 | 35.2 | 35.272 | 0.12 |
2 | 44.8 | 44.642 | -0.26 | |
指针表2 | 1 | 40.2 | 40.069 | -0.22 |
2 | 46.8 | 46.793 | -0.01 |
运动模糊类型 | 序号 | 人工读数/ MPa | 本文方法 读数/MPa | 引用误差/ % | |
---|---|---|---|---|---|
L/像素 | a/° | ||||
20 | 0 | 1 | 0.16 | 0.156 | -0.16 |
2 | 0.75 | 0.752 | 0.08 | ||
20 | 30 | 1 | 0.75 | 0.752 | 0.08 |
2 | 1.38 | 1.375 | -0.20 | ||
20 | 45 | 1 | 1.07 | 1.078 | 0.32 |
2 | 1.38 | 1.375 | -0.20 | ||
20 | 60 | 1 | 1.72 | 1.722 | 0.08 |
2 | 1.90 | 1.904 | 0.16 | ||
30 | 45 | 1 | 0.38 | 0.385 | 0.20 |
2 | 2.25 | 2.245 | -0.20 | ||
40 | 45 | 1 | 1.38 | 1.372 | -0.32 |
2 | 2.38 | 2.372 | -0.32 |
表11 运动模糊的指针式仪表自动读数识别结果
Table 11 Automatic reading recognition results for motion blurred pointer gauges
运动模糊类型 | 序号 | 人工读数/ MPa | 本文方法 读数/MPa | 引用误差/ % | |
---|---|---|---|---|---|
L/像素 | a/° | ||||
20 | 0 | 1 | 0.16 | 0.156 | -0.16 |
2 | 0.75 | 0.752 | 0.08 | ||
20 | 30 | 1 | 0.75 | 0.752 | 0.08 |
2 | 1.38 | 1.375 | -0.20 | ||
20 | 45 | 1 | 1.07 | 1.078 | 0.32 |
2 | 1.38 | 1.375 | -0.20 | ||
20 | 60 | 1 | 1.72 | 1.722 | 0.08 |
2 | 1.90 | 1.904 | 0.16 | ||
30 | 45 | 1 | 0.38 | 0.385 | 0.20 |
2 | 2.25 | 2.245 | -0.20 | ||
40 | 45 | 1 | 1.38 | 1.372 | -0.32 |
2 | 2.38 | 2.372 | -0.32 |
测试 编号 | 图片的读数时间 | |||
---|---|---|---|---|
需运动模糊处理 | 无需运动模糊处理 | |||
30张 | 每张 | 30张 | 每张 | |
1 | 16.776 | 0.559 | 3.839 | 0.128 |
2 | 16.812 | 0.560 | 3.879 | 0.129 |
3 | 16.793 | 0.560 | 3.523 | 0.117 |
4 | 16.912 | 0.564 | 3.835 | 0.127 |
5 | 16.853 | 0.562 | 4.124 | 0.137 |
平均读 数时间/s | 16.829 | 0.561 | 3.840 | 0.128 |
表12 指针式仪表自动读数识别时间测试/s
Table 12 Pointer meter automatic reading recognition time test/s
测试 编号 | 图片的读数时间 | |||
---|---|---|---|---|
需运动模糊处理 | 无需运动模糊处理 | |||
30张 | 每张 | 30张 | 每张 | |
1 | 16.776 | 0.559 | 3.839 | 0.128 |
2 | 16.812 | 0.560 | 3.879 | 0.129 |
3 | 16.793 | 0.560 | 3.523 | 0.117 |
4 | 16.912 | 0.564 | 3.835 | 0.127 |
5 | 16.853 | 0.562 | 4.124 | 0.137 |
平均读 数时间/s | 16.829 | 0.561 | 3.840 | 0.128 |
[1] | 熊国良, 肖文明, 王小明. 基于视觉的指针式仪表检测与识别方法综述[J]. 传感器与微系统, 2020, 39(12): 1-3, 9. |
XIONG G L, XIAO W M, WANG X M. Review of pointer meter detection and recognition method based on vision[J]. Transducer and Microsystem Technologies, 2020, 39(12): 1-3, 9. (in Chinese) | |
[2] | 周泓, 徐海儿, 耿晨歌. 基于HSI模型和Hough变换的指针式汽车仪表自动校验[J]. 浙江大学学报(工学版), 2010, 44(6): 1108-1112. |
ZHOU H, XU H E, GENG C G. Automatic checking of pointer automotive dashboard based on HIS model and Hough transformation[J]. Journal of Zhejiang University (Engineering Science), 2010, 44(6): 1108-1112. (in Chinese) | |
[3] | 黄炎, 李文胜, 麦晓明, 等. 基于一维测量线映射的变电站指针仪表智能识读方法[J]. 广东电力, 2018, 31(12): 80-85. |
HUANG Y, LI W S, MAI X M, et al. Intelligent recognition method for substation pointer meter images based on one-dimensional measuring line mapping[J]. Guangdong Electric Power, 2018, 31(12): 80-85. (in Chinese) | |
[4] |
阎光伟, 刘润泽, 焦润海, 等. 基于改进Cascade RCNN的输电线路防振锤脱落检测方法[J]. 图学学报, 2023, 44(5): 849-860.
DOI |
YAN G W, LIU R Z, JIAO R H, et al. Detection method of dropped anti-vibration hammer for transmission line based on improved Cascade RCNN[J]. Journal of Graphics, 2023, 44(5): 849-860. (in Chinese) | |
[5] | 万吉林, 王慧芳, 管敏渊, 等. 基于Faster R-CNN和U-Net的变电站指针式仪表读数自动识别方法[J]. 电网技术, 2020, 44(8): 3097-3105. |
WAN J L, WANG H F, GUAN M Y, et al. An automatic identification for reading of substation pointer-type meters using faster R-CNN and U-Net[J]. Power System Technology, 2020, 44(8): 3097-3105. (in Chinese) | |
[6] | 陶金, 林文伟, 曾亮, 等. 基于YOLOv4-tiny和Hourglass的指针式仪表读数识别[J]. 电子测量与仪器学报, 2023, 37(5): 1-10. |
TAO J, LIN W W, ZENG L, et al. Pointer meter reading recognition based on YOLOv4-tiny and Hourglass[J]. Journal of Electronic Measurement and Instrumentation, 2023, 37(5): 1-10. (in Chinese) | |
[7] | HOU L Q, WANG S, SUN X P, et al. A pointer meter reading recognition method based on YOLOX and semantic segmentation technology[J]. Measurement, 2023, 218: 113241. |
[8] |
赵振兵, 马迪雅, 石颖, 等. 基于改进YOLOX的变电站仪表外观缺陷检测算法[J]. 图学学报, 2023, 44(5): 937-946.
DOI |
ZHAO Z B, MA D Y, SHI Y, et al. Appearance defect detection algorithm of substation instrument based on improved YOLOX[J]. Journal of Graphics, 2023, 44(5): 937-946. (in Chinese) | |
[9] | UEDA S, SUZUKI K, KANNO J, et al. A two-stage deep learning-based approach for automatic reading of analog meters[C]// The 11th International Conference on Soft Computing and Intelligent Systems and 21st International Symposium on Advanced Intelligent Systems. New York: IEEE Press, 2020: 1-6. |
[10] | 孙晓朋, 侯立群, 渠怀胜. 基于卷积神经网络的渐进式指针表自动读数方法[J]. 传感技术学报, 2021, 34(10): 1326-1333. |
SUN X P, HOU L Q, QU H S. Gradually Automatic reading method for pointer meters using convolutional neural network[J]. Chinese Journal of Sensors and Actuators, 2021, 34(10): 1326-1333. (in Chinese) | |
[11] | ZHANG T Y, SUEN C Y. A fast parallel algorithm for thinning digital patterns[J]. Communications of the ACM, 1984, 27(3): 236-239. |
[12] | 杨武, 胡敏, 常鑫, 等. 改进的DeepLabV3+指针式仪表图像分割算法[J]. 国外电子测量技术, 2024, 43(1): 10-19. |
YANG W, HU M, CHANG X, et al. Improved image segmentation algorithm of DeepLabV3+ pointer meter[J]. Foreign Electronic Measurement Technology, 2024, 43(1): 10-19. (in Chinese) | |
[13] | 朱斌滨, 樊绍胜. 雨雾环境下的变电站指针式仪表识别方法[J]. 激光与光电子学进展, 2021, 58(24): 2410008. |
ZHU B B, FAN S S. Reading method of substation pointer meter in rain-fog environment[J]. Laser & Optoelectronics Progress, 2021, 58(24): 2410008. (in Chinese) | |
[14] | ZHAI X M, PEI D. Research on intelligent reading algorithm of pointer meter based on deep learning[J]. International Core Journal of Engineering, 2022, 8(4): 554-565. |
[15] | 侯卓成, 欧阳华, 胡鑫, 等. 基于深度学习的模糊指针式仪表矫正读数方法[J]. 电子测量技术, 2023, 46(9): 158-165. |
HOU Z C, OUYANG H, HU X, et al. Correction reading method of fuzzy pointer instrument based on deep learning[J]. Electronic Measurement Technology, 2023, 46(9): 158-165. (in Chinese) | |
[16] | MA Y F, JIANG Q, WANG J J, et al. An automatic reading method of pointer instruments[C]// 2017 Chinese Automation Congress. New York: IEEE Press, 2017: 1448-1453. |
[17] | HOU L Q, QU H S. Automatic recognition system of pointer meters based on lightweight CNN and WSNs with on-sensor image processing[J]. Measurement, 2021, 183: 109819. |
[18] | HOU L Q, SUN X P, WANG S. A coarse-fine reading recognition method for pointer meters based on CNN and computer vision[J]. Engineering Research Express, 2022, 4(3): 035046. |
[19] | GAO H J, YI M, YU J Y, et al. Character segmentation-based coarse-fine approach for automobile dashboard detection[J]. IEEE Transactions on Industrial Informatics. 2019, 15(10): 5413-5424. |
[20] | LI Z, ZHOU Y S, SHENG Q H, et al. A high-robust automatic reading algorithm of pointer meters based on text detection[J]. Sensors, 2020, 20(20): 5946. |
[21] | WANG C Y, BOCHKOVSKIY A, LIAO H Y M. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]// 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2023: 7464-7475. |
[22] | YANG X, YAN J C. Arbitrary-oriented object detection with circular smooth label[C]// The 16th European Conference on Computer Vision. Cham: Springer, 2020: 677-694. |
[23] | OUYANG D L, HE S, ZHANG G Z, et al. Efficient multi-scale attention module with cross-spatial learning[C]// 2023 IEEE International Conference on Acoustics, Speech and Signal Processing. New York: IEEE Press, 2023: 1-5. |
[24] | LEE J, PARK S, MO S, et al. Layer-adaptive sparsity for the magnitude-based pruning[EB/OL]. (2021-05-09) [2024-05-23]. https://arxiv.org/abs/2010.07611. |
[25] | YANG X, YANG J R, YAN J C, et al. SCRDeT: towards more robust detection for small, cluttered and rotated objects[C]// 2019 IEEE/CVF International Conference on Computer Vision. New York: IEEE Press, 2019: 8231-8240. |
[26] | CHO S J, JI S W, HONG J P, et al. Rethinking coarse-to-fine approach in single image deblurring[C]// 2021 IEEE/CVF International Conference on Computer Vision. New York: IEEE Press, 2021: 4621-4630. |
[27] | ZAMIR S W, ARORA A, KHAN S, et al. Restormer: efficient transformer for high-resolution image restoration[C]// 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2022: 5718-5729. |
[28] | NEWELL A, YANG K Y, DENG J. Stacked hourglass networks for human pose estimation[C]// The 14th European Conference on Computer Vision. Cham: Springer, 2016: 483-499. |
[29] | LIU Y C, SHAO Z R, TENG Y Y, et al. NAM: normalization-based attention module[EB/OL]. (2021-11-24) [2024-05-23]. https://arxiv.org/abs/2111.12419. |
[30] | CUI Y N, TAO Y, BING Z S, et al. Selective frequency network for image restoration[EB/OL]. [2024-05-23]. https://openreview.net/forum?id=tyZ1ChGZIKO. |
[31] | 半片青柠. 工业场景下指针表数据集_语义分割做指针识别资源-CSDN文库[EB/OL]. (2024-05-11) [2024-05-23]. https://download.csdn.net/download/sinat_40587853/87453496? |
Half a Slice of Lime. Pointer table dataset for industrial scenarios_semantic segmentation for pointer recognition resources - CSDN library[EB/OL]. (2024-05-11) [2024-05-23]. https://download.csdn.net/download/sinat_40587853/87453496?. (in Chinese) |
[1] | 闫建红, 冉同霄. 基于YOLOv8的轻量化无人机图像目标检测算法[J]. 图学学报, 2024, 45(6): 1328-1337. |
[2] | 曾志超, 徐玥, 王景玉, 叶元龙, 黄志开, 王欢. 基于SOE-YOLO轻量化的水面目标检测算法[J]. 图学学报, 2024, 45(4): 736-744. |
[3] | 朱强军, 胡斌, 汪慧兰, 王杨. 基于轻量化YOLOv8s交通标志的检测[J]. 图学学报, 2024, 45(3): 422-432. |
[4] | 苑朝, 赵亚冬, 张耀, 王嘉璇, 徐大伟, 翟永杰, 朱松松. 基于YOLO轻量化的多模态行人检测算法[J]. 图学学报, 2024, 45(1): 35-46. |
[5] | 翟永杰, 赵晓瑜, 王璐瑶, 王亚茹, 宋晓轲, 朱浩硕. IDD-YOLOv7:一种用于输电线路绝缘子多缺陷的轻量化检测方法[J]. 图学学报, 2024, 45(1): 90-101. |
[6] | 皮骏, 牛厚兴, 高志云. 融合CA-BiFPN的轻量化人体姿态估计算法[J]. 图学学报, 2023, 44(5): 868-878. |
[7] | 熊举举, 徐杨, 范润泽, 孙少聪. 基于轻量化视觉Transformer的花卉识别[J]. 图学学报, 2023, 44(2): 271-279. |
[8] | 皮骏, 刘宇恒, 李久昊. 基于YOLOv5s的轻量化森林火灾检测算法研究[J]. 图学学报, 2023, 44(1): 26-32. |
[9] | 张运波, 易鹏飞, 周东生, 张强, 魏小鹏. 深度可分离卷积和标准卷积相结合的高效行人检测器[J]. 图学学报, 2022, 43(2): 230-238. |
[10] | 张芳兰, 刘龙吉, 姚宛彤. 面向关键用户需求的踝足矫形器定制化设计方法[J]. 图学学报, 2021, 42(5): 841-848. |
[11] | 肖文磊, 王志明, 王世平, 赵 罡. 基于周期性晶格的 3D 打印模型轻量化方法[J]. 图学学报, 2021, 42(2): 263-270. |
[12] | 董 健 1, 董玉德 1, 刘 福 1, 王 帅 1, 时晓蕾 2, 陶高周 2 . 基于复杂装配体特征抑制的轻量化方法[J]. 图学学报, 2019, 40(5): 968-975. |
[13] | 郭思怡, 陈永锋. 建筑运维阶段信息模型的轻量化方法[J]. 图学学报, 2018, 39(1): 123-128. |
[14] | 王晓斌, 宁 涛, 王 可. 3DXML文件格式解析及应用[J]. 图学学报, 2010, 31(2): 33-37. |
[15] | 谢 峰, 沈维蕾, 林巨广. C型压力机机身的轻量化设计[J]. 图学学报, 2010, 31(1): 13-19. |
阅读次数 | ||||||
全文 |
|
|||||
摘要 |
|
|||||