【读论文】稀疏特征

最新推荐文章于 2024-07-05 17:05:14 发布

Vine955

最新推荐文章于 2024-07-05 17:05:14 发布

阅读量197

点赞数

分类专栏：机器学习

本文链接：https://blog.csdn.net/weixin_40389065/article/details/116588632

版权

2 篇文章 0 订阅

订阅专栏

Pavement crack detection and recognition using the architecture of segNet

Crack detection is a typical semantic segmentation task; 语义分割问题
We propose an encoder-decoder structural model with a fully convolutional neural network, namely, PCSN, by referring to SegNet. 提出PCSN，基于SegNet。
A crack dataset of images containing complex crack textures is constructed by manual pixelwise annotation.

Pavement and bridge defects consist of internal invisible defects and surface visible defects. 包含内部不可见的裂痕以及表面可见的裂痕。
Surface visible crack defects are a long-standing problem and affect public safety. 引发安全问题
However, due to the complex road environment, it is difficult to detect pavement and bridge surface cracks using such a processing flow [6]. 道路复杂性，传统方法不可行。
The process of detecting cracks from images can be seen as an object detection, classification and localization problem.
Therefore, deep learning based model can be used to detect pavement and bridge surface crack defects. 考虑深度学习方法。
Our main strategy, which is based on an analysis of previous research on crack inspection, is to improve the performance of pavement crack detection method by integrating the feature of low-resolution input sampling. 针对低分辨率的图片。
部分数据来源.

To simplify the structure of crack detection models and to reduce the number of internal parameters, pixel-wise detection approaches, such as FCN and SegNet, are proposed by researchers due to the excellent advantages of simple structure and fewer weight parameters.

解释PCSN为什么需要非常多的数据（labeled images）。
如何获取images。
images组成以及打标。
images尺寸。
调整images的角度、亮度、对比度、色彩等参数，以扩充数据。（thereby enabling PCSN to learn a wider range of crack patterns. ）
训练集、验证集、测试集的分配情况。

In this paper, the key point is how to define the issue of pavement and bridge crack that exists on concrete or asphalt pavement and bridge deck and how to adopt full advantages of CNN, FCN and SegNet. 如何识别混凝土或者柏油马路上面的crack。