| 基于改进Mask R⁃CNN的航空发动机保险丝实例分割方法 |
| Aero⁃engine lockwire instance segmentation method based on improved Mask R⁃CNN |
| |
| HTML View Full Text View/Add Comment Download reader |
| 中文摘要: |
| 针对成像背景复杂、光照不均、目标区域占比小等因素导致的航空发动机保险丝识别精度低的问题，提出一种改进的基于掩模区域的卷积神经网络（Mask Region?based Convolutional Neural Network， Mask R?CNN）保险丝实例分割模型。首先分别对保险丝图像的R、G、B三个通道进行不同程度的伽马校正，转化得到伪彩色图像，同时增强对比度；然后，针对保险丝的细长曲线几何特征，将动态蛇形卷积融入Mask R?CNN的骨干网络Resnet中，使得网络在特征提取时自适应地聚焦细长弯曲的局部结构；最后在特征融合阶段引入卷积注意力模块（Convolution Block Attention Module， CBAM），保留小目标浅层特征，从而提高网络对小目标的感知能力。实验结果表明，改进后的模型掩码AAP50达到了82.54%，较基础模型提升了5.83%，为航空发动机保险丝数字化、智能化检测提供了有力支撑。 |
| 英文摘要: |
| In order to solve the problem of low recognition accuracy of aero?engine lockwire caused by factors of the complex background, uneven illumination and small percentage of the target region, this paper proposes an improved mask region?based convolutional neural network (Mask R?CNN) model for lockwire instance segmentation. Firstly, the gamma corrections of R, G and B channels with different degrees were carried out to transform the lockwire image into pseudo?color image and enhance the contrast. Then, the dynamic snake?shaped convolution was incorporated into Resnet, the backbone network of Mask R?CNN, to make the network to adaptively focus on the slender and curved local structure during feature extraction. Then, based on the geometric features of the fuse's slender curve, dynamic snake convolution was integrated into the backbone network Resnet of Mask R?CNN, allowing the network to adaptively focus on the local structure of the slender curve during feature extraction. Finally, the CBAM attention mechanism was introduced in the feature fusion phase to retain the shallow features of small target, so as to improve the perception ability of the network on small target. The experimental results showed that the AAP50 of the improved module mask reached 82.54%, which was improved by 5.83% compared to basic mode. This study provides strong support for digital and intelligent detection of aero?engine lockwire. |
|
| 中文关键词: 航空发动机保险丝 基于掩模区域的卷积神经网络 实例分割 动态蛇形卷积 特征提取 卷积注意力模块 深度学习 |
| 英文关键词:aero⁃engine lockwire Mask R⁃CNN instance segmentation dynamic snake⁃shaped convolution feature extraction CBAM deep learning |
| 基金项目: |
| DOI：10.11823/j.issn.1674-5795.2025.01.07 |
| Close |
