基于不确定性增强的RGB-IR双波段图像语义分割算法
Semantic segmentation algorithm for RGB-IR dual-band images based on uncertainty enhancement
陈定律1
作者信息
- 1. 浙江理工大学计算机科学与技术学院,浙江 杭州 310018
- 折叠
摘要
语义分割广泛应用于机器人、医学成像和自动驾驶等领域,但当前语义分割主要针对可见光图像.可见光图像在光照不足或天气差的情况下成像效果较差,而红外图像受光照影响较小.因此,将可见光图像和红外图像联合使用可以提升模型的鲁棒性.通过预测前景轮廓的不确定性并将其作为注意力机制,可以有效地提高模型在前景物体和边缘轮廓部分的分割能力.本文模型在公开数据库上取得了57.2的分割精度,综合性能优秀.
Abstract
Semantic segmentation is widely used in robotics,medical imaging,autonomous driving and other fields.However,the existing methods mainly focus on visible(RGB)images which have low quality under insufficient illumination or bad weather conditions.The infrared(IR)images are less affected by such situations.Therefore,the combination of RGB and IR images can improve the robustness of the model.By predicting the uncertainty of the foreground edge and using it as an attention mechanism,the model's segmentation ability in the foreground object and edge parts can be effectively improved.The proposed model achieves a segmentation accuracy of 57.2 on a public dataset with excellent comprehensive performance.
关键词
语义分割/不确定性/注意力机制/红外图像Key words
semantic segmentation/uncertainty/attention mechanism/IR image引用本文复制引用
出版年
2023
NETL
NSTL
维普
万方数据