1 中国科学院水利部成都山地灾害与环境研究所，成都 610299
2 中国科学院大学，北京 100049
冰湖溃决洪水是一种严重的山地自然灾害，威胁着中国高寒区的居民及铁路公路等重要基础设施，自动高效的冰湖遥感制图方法是冰湖灾害评估、监测预警的基础，然而现有自动制图方法在实际冰湖提取应用上难以达到传统人工和半自动冰湖提取方法上的精度，仍需进一步提高。文章在原生U-Net模型基础上，在各桥连接部分融合极化自注意力机制，将输入影像特征分别在空间和通道层保持高分辨率，并通过非线性合成输出细腻的特征，构建了一种改进的U-Net冰湖遥感深度学习制图方法，并将其成功应用在高原铁路关键区。研究结果表明：1）与PSPNet、DeepLabV3+、原生U-Net三种经典模型相比，改进模型在冰湖预测数据集上的各项指标上都有提升，精确率、召回率、交并比和F1值分别达到了0.972 5、0.966 5、0.940 8和0.969 4，相较于原生U-Net网络，精确度、召回率、交并比和F1值分别提高了5.01%、6.05%、10.73%和5.53%；2）基于Landsat-8卫星遥感数据，应用改进模型完成了2013—2022年帕隆藏布和易贡藏布案例区冰湖信息自动高效提取，如2020年冰湖总体精度为98.16%，与参照数据的重叠度达到96.66%，提取的精度满足冰湖灾害评估和监测预警研究需求，可用于铁路等重大工程沿线冰湖灾害防治的实践。
遥感监测 冰湖灾害 深度学习 自注意力机制 U-Net卷积神经网络 remote sensing monitoring glacial lake disaster deep learning self-attention mechanism U-Net
精准高效地从高分辨率遥感影像中提取建筑物信息对国土规划和地图制图意义重大，近年来基于卷积神经网络进行建筑物信息提取已经取得了很大的进展，然而在处理高分辨率遥感影像时仍存在影像的高级语义特征利用不够充分，难以获得细节丰富高精度分割影像的问题。文章针对以上问题提出了一种用于建筑物自动提取的深度学习网络结构空洞空间与通道感知网络(Atrous Space and Channel Perception Network，ASCP-Net）。该模型将空洞空间金子塔池化（Atrous Spatial Pyramid Pooling, ASPP）和空间与通道注意力 （Spatial and Channel Attention, SCA）模块融入到编码器-解码器结构中，通过ASPP模块来捕获和聚合多尺度上下文信息，采用SCA模块选择性增强特定位置和通道中更有用的信息，并将高低层特征信息输入解码网络完成建筑物信息的高效提取。在WHU建筑数据集（WHU Building Dataset）上进行实验，结果表明：文章提出的方法总体精度和F1评分分别达到了97.4%和94.6%，相比其他模型能够获得更清晰的建筑物边界，尤其对图像边缘不完整建筑的提取效果较好，有效提升了建筑物提取的精度和完整性。
高分辨率遥感影像 双注意力机制 空洞卷积 建筑物提取 high-resolution remote sensing images dual attention mechanism atrous convolution building extraction
1 黔西南州自然资源管理服务中心，兴义 562400
2 贵州大学矿业学院，贵阳 550025
针对传统喀斯特地区裸岩提取方法成本高、精度低的问题，文章构建了一种基于改进DeepLabV3+的裸岩提取方法。该方法首先在编码器中用CA-DC-MobileNetV3替换DeepLabV3+骨干网络Xception进行特征提取，很大程度上减少了模型的参数量；其次，将编码器提取的特征通过特征金字塔网络和坐标注意力机制进行加强特征提取，以获取更多小目标信息并减少图像细节损失；最后在空洞空间金字塔池化模块将不同空洞率的卷积层进行特征融合，提高信息的利用率。研究结果表明：文章方法在不同场景裸岩提取任务中表现最好，模型参数量约为DeepLabV3+的1/13，交并比、F1分数分别为72.46%、84.03%，上述2个指标相比于DeepLabV3+模型分别提高了4.62和3.19个百分点，并优于其余常用语义分割模型，提高了裸岩提取精度。
裸岩提取 深度学习 语义分割 坐标注意力机制 bare rock extraction deep learning semantic segmentation coordinate attention mechanism
Author Affiliations
Abstract
1 School of Biomedical Engineering (Suzhou), Division of Life Sciences and Medicine, University of Science and Technology of China, Hefei, Anhui 230026, P. R. China
2 Suzhou Institute of Biomedical Engineering and Technology, Chinese Academy of Sciences, Suzhou, Jiangsu 215163, P. R. China
3 School of Physical Science and Technology, Suzhou University of Science and Technology, Suzhou, Jiangsu 215009, P. R. China
Structured illumination microscopy (SIM) is a popular and powerful super-resolution (SR) technique in biomedical research. However, the conventional reconstruction algorithm for SIM heavily relies on the accurate prior knowledge of illumination patterns and signal-to-noise ratio (SNR) of raw images. To obtain high-quality SR images, several raw images need to be captured under high fluorescence level, which further restricts SIM’s temporal resolution and its applications. Deep learning (DL) is a data-driven technology that has been used to expand the limits of optical microscopy. In this study, we propose a deep neural network based on multi-level wavelet and attention mechanism (MWAM) for SIM. Our results show that the MWAM network can extract high-frequency information contained in SIM raw images and accurately integrate it into the output image, resulting in superior SR images compared to those generated using wide-field images as input data. We also demonstrate that the number of SIM raw images can be reduced to three, with one image in each illumination orientation, to achieve the optimal tradeoff between temporal and spatial resolution. Furthermore, our MWAM network exhibits superior reconstruction ability on low-SNR images compared to conventional SIM algorithms. We have also analyzed the adaptability of this network on other biological samples and successfully applied the pretrained model to other SIM systems.
Super-resolution reconstruction multi-level wavelet packet transform residual channel attention selective kernel attention Journal of Innovative Optical Health Sciences
2024, 17(2): 2350015
