[1]王 静*,祝 鹏.doi: 10.3969/j.issn.1001-3849.2025.11.013高亮异形电镀件表面斑点缺陷智能识别研究[J].电镀与精饰,2025,(11):93-99.
　Wang Jing*,Zhu Peng.Research on intelligent recognition of surface spot defects on high brightness irregular electroplated parts[J].Plating & Finishing,2025,(11):93-99.
点击复制

doi: 10.3969/j.issn.1001-3849.2025.11.013高亮异形电镀件表面斑点缺陷智能识别研究()

《电镀与精饰》[ISSN:1001-3849/CN:12-1096/TG]

卷:
期数:
2025年11
页码:
93-99
栏目:
出版日期:
2025-11-30

文章信息/Info

Title:
Research on intelligent recognition of surface spot defects on high brightness irregular electroplated parts
作者:
王 静1*; 祝 鹏2
(1. 赤峰学院 数学与计算机科学学院,内蒙古 赤峰 024000 ;2. 内蒙古农业大学 计算机技术与信息管理系,内蒙古 呼和浩特 010018)
Author(s):
Wang Jing1*;  Zhu Peng2
(1. College of Mathematics and Computer Science, Chifeng University, Chifeng 024000, China; 2. Department of Computer Technology and Information Management, Inner Mongolia Agricultural University, Hohhot 010018, China)
关键词:
高亮异形电镀件; 斑点缺陷识别; Retinex增强; 蝙蝠算法; Res-UNet网络
Keywords:
high gloss shaped electroplated parts;  spot defect recognition;  retinex enhancement;  bat algorithm;  res unet network
分类号:
TP391.4
文献标志码:
A
摘要:
高亮异形电镀件表面呈现出高度不规则的几何形貌，这种复杂的微观结构特性使其在多变的光照条件下极易产生反射光斑现象。这些反射光斑区域与周边低亮度区域之间形成了极为悬殊的亮度差异，进而引发图像失真问题，降低斑点特征的识别精度。为此，提出高亮异形电镀件表面斑点缺陷智能识别研究。将原始电镀件表面图像划分为Retinex增强层、亮度增强层和细节突出层，获得增强后的电镀件图像。采用蝙蝠算法对Res-UNet网络的特征提取层数进行优化后，将增强后的电镀件图像输入到改进后的Res-UNet网络中，该网络能够自动学习电镀件表面斑点缺陷的特征表示，并精准地将其从背景中分割出来，从而实现斑点缺陷的精确识别。实验结果表明，该方法能够有效消除光晕与噪点，准确识别电镀件表面的斑点缺陷。
Abstract:
The surface of the high gloss irregular electroplated parts presents a highly irregular geometric morphology, and this complex microstructural characteristic makes it prone to reflection spot phenomenon under variable lighting conditions. These reflected light spot areas form a significant brightness difference with the surrounding low brightness areas, which leads to image distortion and reduces the recognition accuracy of spot features. Therefore, a research on intelligent recognition of surface spot defects on high brightness irregular electroplated parts is proposed. Divide the original surface image of the electroplated part into Retinex enhancement layer, brightness enhancement layer, and detail highlighting layer to obtain the enhanced electroplated part image. After optimizing the feature extraction layers of the Res UNet network using the bat algorithm, the enhanced electroplated image is input into the improved Res UNet network, which can automatically learn the feature representation of surface spot defects on electroplated parts and accurately segment them from the background, thereby achieving precise recognition of spot defects. The experimental results show that this method can effectively eliminate halo and noise, and accurately identify spot defects on the surface of electroplated parts

参考文献/References:

[1].张维召, 王静. 汽车后门外板电泳前基板团斑缺陷分析[J]. 电镀与涂饰, 2023, 42(8): 41-44.
[2].张斌, 郭宇会, 杨小平, 等. 辊径对Cu-Fe-P铜合金轧制带材表面微观缺陷及性能的影响[J]. 锻压技术, 2024, 49(9): 122-128.
[3].付泓豪, 张岱凌, 江澜, 等. 基于图像高频增强的塑料齿轮黑点缺陷高精度检测方法[J]. 工具技术, 2023, 57(4): 132-136.
[4].胡广华, 涂千禧. 基于光度立体和双流特征融合网络的工业产品表面缺陷检测方法[J]. 华南理工大学学报(自然科学版), 2024, 52(10): 112-123.
[5].Teng W Z, Zhang Y J, Zhang H G, et al. Surface defect detection of steel based on improved YOLOv7 model[J]. Metalurgija, 2024, 63(3-4): 399-402.
[6].Tang B, Chen L, Sun Z K. Review of surface defect detec-tion of steel products based on machine vision[J]. IET Im-age Processing, 2023, 17(2): 303-322.
[7].李梦蕊, 柳晓鸣, 常婧. 基于颜色衰减先验的自适应Retinex去雾算法[J]. 计算机仿真, 2021, 38(1): 354-358.
[8].赵征鹏, 李俊钢, 普园媛. 基于卷积神经网络的Retinex低照度图像增强[J]. 计算机科学, 2022, 49(6): 199-209.
[9].张延军, 夏黎明. 基于改进加权引导滤波的煤矿井下图像除雾算法研究[J]. 矿业研究与开发, 2023, 43(5): 203-210.
[10].武凌霄, 康家银, 姬云翔. NSST域下基于引导滤波与稀疏表示的红外与可见光图像融合[J]. 红外技术, 2023, 45(9): 915-924.
[11].苏波, 李超, 王莉. 基于多权重融合策略的Retinex矿井图像增强算法[J]. 煤炭学报, 2023, 48(S2): 813-822.
[12].任雨霞, 曾理. 基于Retinex模型和GTV的铁路货车铸件DR图像增强[J]. 铁道科学与工程学报, 2023, 20(2): 706-713.
[13].陆涛. 基于统计特征分类耦合自适应Gamma校正的图像增强算法[J]. 电子测量与仪器学报, 2020, 34(6): 154-162.
[14].廖娟, 陈民慧, 汪鹞, 等. 基于双重Gamma校正的秧苗图像增强算法[J]. 江苏农业学报, 2020, 36(6): 1411-1418.
[15].吕石磊, 范仁杰, 李震, 等. 基于改进蝙蝠算法和圆柱坐标系的农业无人机航迹规划[J]. 农业机械学报, 2023, 54(1): 20-29, 63.
[16].赵太飞, 容开新, 王一琼, 等. 改进蝙蝠算法的紫外光引导无人机路径规划[J]. 激光技术, 2023, 47(5): 678-685.
[17].姚艳, 高飞翎, 周子旺. 基于改进蝙蝠算法的含分布式电源配电网无功优化[J]. 电气传动, 2023, 53(6): 46-52.
[18].许天恩, 周怀来, 刘兴业, 等. 基于Res-Unet与迁移学习的地震相识别[J]. 地球物理学进展, 2024, 39(1): 319-333.
[19].吕少云, 李佳田, 阿晓荟, 等. Res＿ASPP＿UNet++:结合分离卷积与空洞金字塔的遥感影像建筑物提取网络[J].遥感学报, 2023, 27(2): 502-519.
[20].廖小华, 陈念年, 蒋勇, 等. 改进的卷积神经网络红外图像超分辨率算法[J]. 红外技术, 2020, 42(1): 75-80.

更新日期/Last Update: 2025-11-20