| 基于CNN的公路路网自动提取方法研究 |
 点此下载全文 |
| 引用本文：昌宏哲1,杜红静1，袁 洋2.基于CNN的公路路网自动提取方法研究[J].计算技术与自动化,2022,(2):164-167 |
| 摘要点击次数: 338 |
| 全文下载次数: 0 |
|
|
| 中文摘要:为提高深度学习网络中公路提取性能，提出了一种基于CNN框架的公路路网自动提取方法。为进一步提高模型训练效果，设计了一种考虑公路结构的损失函数。计算损失时不仅考虑每个像素的重要程度，同时考虑公路的全局结构。通过仿真分析，所提方法准确率达到92.4%，较传统CNN方法提高13%。 |
| 中文关键词:CNN 公路路网 自动提取 交叉熵损失 |
| |
| Research on Automatic Extraction Method of Highway Network Based on CNN |
|
|
| Abstract:In order to improve the performance of road network extraction in deep learning network, an automatic extraction method of road network based on CNN framework is proposed. In order to further improve the training effect of the model, a loss function considering the highway structure is designed too. When calculating the loss, not only the importance of each pixel is considered, but also the global structure of highway is considered. Through simulation analysis, the accuracy of the proposed method reaches 92.4%, which is 13% higher than that of the traditional CNN method. |
| keywords:CNN road network automatic extraction cross entropy loss |
| 查看全文 查看/发表评论 下载pdf阅读器 |
