按文献类别分组
按栏目分组
关键词
基于多尺度U-Net与Transformer特征融合的航空遥感图像飞机检测方法
作者: 张善文     邵彧     李萍     令伟锋   来源: 弹箭与制导学报 年份: 2024 文献类型 : 期刊 关键词: Transformer   Net与Transformer   航空遥感图像飞机检测   多尺度U   Net  
描述: 航空遥感图像(ARSI)飞机检测一直是一个重要且具有挑战性的课题。针对现有ARSI飞机检测方法(ARSIAD)检测目标的边缘模糊、小目标的检测精度低、没有充分利用ARSI的全局上下文信息等问题,提出一种基于多尺度U-Net与Transformer (MSU-Trans)特征融合的ARSIAD方法。通过多尺度卷积模块Inception提取ARSI中多样性目标的分类特征,通过Transformer增强模型的全局语义检测性能,通过特征融合模块整合高层和低层特征,得到航空目标图像完整的边缘和纹理特征。该模型结合多尺度U-Net较强的局部特征提取能力和Transformer较强的全局上下文依存关系提取能力,进而提高MSU-Trans的整体检测性能。在ARSI集上的试验表明,与U-Net、多尺度U-Net、注意力U-Nets相比,MSU-Trans具有较高的检测精度,精度超过95%,该方法为ARSIAD提供一定的技术支撑。
低转速航空发动机滚动轴承故障深度异常检测方法
作者: 康玉祥     陈果     盛嘉玖     王浩     尉询楷   来源: 振动与冲击 年份: 2024 文献类型 : 期刊 关键词: 滚动轴承   航空发动机   Transformer   深度异常检测   低转速  
描述: 结果作为网络的输入,且仅依靠正常数据完成模型的训练。最后,在带机匣的航空发动机转子试验器和某型真实的航空发动机上分别进行了试验验证。结果表明,所提方法能够准确的实现对低转速滚动轴承故障的检测,且检测精度分别为93%和100%,充分表明该方法具有很好的异常检测能力及应用价值。
基于概率稀疏自注意力的航空发动机剩余寿命预测
作者: 王欣     黄佳琪     许雅玺   来源: 科学技术与工程 年份: 2024 文献类型 : 期刊 关键词: 航空发动机   Transformer   深度学习   概率稀疏自注意力   剩余寿命预测  
描述: 航空发动机剩余寿命预测对其健康管理具有重要意义,针对长序列、多维度的航空发动机监测参数,提出一种基于概率稀疏自注意力(ProbSparse Self-Attention)的Transformer模型以实现航空发动机剩余寿命的准确预测。用ProbSparse Self-Attention取代原始Transformer中的常规自注意力机制,使得模型更关注时间序列中重要的时间节点,大幅缩减时间维度,减小了时间和空间复杂度;通过注意力层整合后的信息,进一步通过前馈神经网络层和卷积层,提取传感器的空间特征,编码层之间通过扩张因果卷积相连接,扩大了感受野,提高了模型对长序列信息的捕获能力。在新公开的N-CMAPSS数据集上验证算法,实验结果表明,相比于实验中的对比模型,所提模型的RMSE和Score值均有提升,推理速度也优于其他模型。
一种基于Transformer编码器与LSTM的飞机轨迹预测方法
作者: 李明阳     鲁之君     曹东晶     曹世翔   来源: 航天返回与遥感 年份: 2024 文献类型 : 期刊 关键词: Transformer   Encoder   轨迹预测   Transformer编码器   飞机目标   神经网络   LSTM模型  
描述: 为了解决飞机目标机动数据集缺失的问题,文章利用运动学建模生成了丰富的轨迹数据集,为网络训练提供了必要的数据支持。针对现阶段轨迹预测运动学模型建立困难及时序预测方法难以提取时空特征的问题,提出了一种结合Transformer编码器和长短期记忆网络(Long Short Term Memory,LSTM)的飞机目标轨迹预测方法,即Transformer-Encoder-LSTM模型。新模型可同时提供LSTM和Transformer编码器模块的补充历史信息和基于注意力的信息表示,提高了模型能力。通过与一些经典神经网络模型进行对比分析,发现在数据集上,新方法的平均位移误差减小到0.22,显著优于CNN-LSTMAttention模型的0.35。相比其他网络,该算法能够提取复杂轨迹中的隐藏特征,在面对飞机连续转弯、大机动转弯的复杂轨迹时,能够保证模型的鲁棒性,提升了对于复杂轨迹预测的准确性。
基于航空图像的目标检测算法Trans_YOLOv5
作者: 文青     伍欣     敖斌     李宽     殷建平   来源: 计算机技术与发展 年份: 2024 文献类型 : 期刊 关键词: 航空图像   Transformer   YOLOv5   圆形平滑标签   小目标检测   Swin  
描述: 能力,使网络模型更加关注于待检测的目标对象。在DOTAv2.0航空图像数据集上的实验结果验证了所提方法的有效性,检测结果达到60.98%mAP,与原YOLOv5算法检测结果相比提高10.85百分点,与官网公布的竞赛最佳结果相比提高2.01百分点。
改进的YOLOv5s遥感影像机场场面飞机小目标识别
作者: 张新君     赵春霖   来源: 电光与控制 年份: 2024 文献类型 : 期刊 关键词: 坐标注意力机制   遥感影像   Transformer   YOLOv5s   小目标检测   Swin  
描述: 小目标检测识别测试实验,改进后的YOLOv5s网络的mAP值为0.837 5,比YOLOv5s网络模型提高了0.022 5。实验结果表明,改进后的YOLOv5s网络模型对比YOLO系列网络和EfficientDet模型有效地提高了识别准确率、召回率以及mAP值,并且在训练时间上也比YOLOv5s减少了1/12。
EGMT-CD: Edge-Guided Multimodal Transformers Change Detection from Satellite and Aerial Images
作者: Yunfan   Xiang     Xiangyu   Tian     Yue   Xu     Xiaokun   Guan     Zhengchao   Chen   来源: Remote Sensing 年份: 2023 文献类型 : 期刊 关键词: sensing   detection   images   edge   Transformer   feature   change   remote   alignment   heterogeneous  
描述: Change detection from heterogeneous satellite and aerial images plays a progressively important role in many fields, including disaster assessment, urban construction, and land use monitoring. Currently, researchers have mainly devoted their attention to change detection using homologous image pairs and achieved many remarkable results. It is sometimes necessary to use heterogeneous images for change detection in practical scenarios due to missing images, emergency situations, and cloud and fog occlusion. However, heterogeneous change detection still faces great challenges, especially using satellite and aerial images. The main challenges in satellite and aerial image change detection are related to the resolution gap and blurred edge. Previous studies used interpolation or shallow feature alignment before traditional homologous change detection methods, which ignored the high-level feature interaction and edge information. Therefore, we propose a new heterogeneous change detection model based on multimodal transformers combined with edge guidance. In order to alleviate the resolution gap between satellite and aerial images, we design an improved spatially aligned transformer (SP-T) with a sub-pixel module to align the satellite features to the same size of the aerial ones supervised by a token loss. Moreover, we introduce an edge detection branch to guide change features using the object edge with an auxiliary edge-change loss. Finally, we conduct considerable experiments to verify the effectiveness and superiority of our proposed model (EGMT-CD) on a new satellite–aerial heterogeneous change dataset, named SACD. The experiments show that our method (EGMT-CD) outperforms many previously superior change detection methods and fully demonstrates its potential in heterogeneous change detection from satellite–aerial images.
EGMT-CD: Edge-Guided Multimodal Transformers Change Detection from Satellite and Aerial Images
作者: Yunfan   Xiang     Xiangyu   Tian     Yue   Xu     Xiaokun   Guan     Zhengchao   Chen   来源: Remote Sensing 年份: 2023 文献类型 : 期刊 关键词: sensing   detection   images   edge   Transformer   feature   change   remote   alignment   heterogeneous  
描述: Change detection from heterogeneous satellite and aerial images plays a progressively important role in many fields, including disaster assessment, urban construction, and land use monitoring. Currently, researchers have mainly devoted their attention to change detection using homologous image pairs and achieved many remarkable results. It is sometimes necessary to use heterogeneous images for change detection in practical scenarios due to missing images, emergency situations, and cloud and fog occlusion. However, heterogeneous change detection still faces great challenges, especially using satellite and aerial images. The main challenges in satellite and aerial image change detection are related to the resolution gap and blurred edge. Previous studies used interpolation or shallow feature alignment before traditional homologous change detection methods, which ignored the high-level feature interaction and edge information. Therefore, we propose a new heterogeneous change detection model based on multimodal transformers combined with edge guidance. In order to alleviate the resolution gap between satellite and aerial images, we design an improved spatially aligned transformer (SP-T) with a sub-pixel module to align the satellite features to the same size of the aerial ones supervised by a token loss. Moreover, we introduce an edge detection branch to guide change features using the object edge with an auxiliary edge-change loss. Finally, we conduct considerable experiments to verify the effectiveness and superiority of our proposed model (EGMT-CD) on a new satellite–aerial heterogeneous change dataset, named SACD. The experiments show that our method (EGMT-CD) outperforms many previously superior change detection methods and fully demonstrates its potential in heterogeneous change detection from satellite–aerial images.
< 1 2 3
Rss订阅