微动脉瘤是糖尿病视网膜病变的初期症状,消除该病灶可在早期非常有效地预防糖尿病视网膜病变。但由于视网膜结构复杂,同时眼底图像的成像由于患者、环境、采集设备等因素的不同会存在不同的亮度和对比度,现有的微动脉瘤检测算法难以实...微动脉瘤是糖尿病视网膜病变的初期症状,消除该病灶可在早期非常有效地预防糖尿病视网膜病变。但由于视网膜结构复杂,同时眼底图像的成像由于患者、环境、采集设备等因素的不同会存在不同的亮度和对比度,现有的微动脉瘤检测算法难以实现该病灶的精确检测和定位,为此本文提出嵌入SENet(squeeze-andexcitation networks)的改进YOLO(you only look once)v4自动检测算法。该算法在YOLOv4网络基础上,首先通过使用一种改进的快速模糊C均值聚类算法对目标样本进行先验框参数优化,以提高先验框与特征图的匹配度;然后,在主干网络嵌入SENet模块,通过强化关键信息,抑制背景信息,提高微动脉瘤的置信度;此外,还在网络颈部增加空间金字塔池化结构以增强主干网络输出特征的接受域,从而有助于分离出重要的上下文信息;最后,在Kaggle数据集上进行模型验证,并与其他方法进行对比。实验结果表明,与其他各种结构的YOLOv4网络模型相比,所提出的嵌入SENet的改进YOLOv4网络模型能显著提高检测结果(与原始YOLOv4相比Fscore提升了12.68%);与其他网络模型以及方法相比,所提出的嵌入SENet的改进YOLOv4网络模型的自动检测精度明显更优,且可实现精准定位。故本文所提出的嵌入SENet的改进YOLOv4算法性能较优,能准确、有效地检测并定位出眼底图像中的微动脉瘤。展开更多
语义分割和深度估计任务是对图像像素级分类的研究,是两个高度相关的任务。从共享特征学习和特征交互融合两个角度出发,提出两个不同的多任务学习架构,即基于压缩激励模块(Squeeze-and-Excitation,SE)和金字塔池化的多任务学习网络(Mult...语义分割和深度估计任务是对图像像素级分类的研究,是两个高度相关的任务。从共享特征学习和特征交互融合两个角度出发,提出两个不同的多任务学习架构,即基于压缩激励模块(Squeeze-and-Excitation,SE)和金字塔池化的多任务学习网络(Multi-task Learning with SE and Pyramid Pooling,MTL_SPP),以及基于压缩激励和可选择权重(Selective Weight,SW)的多任务学习网络(Multi-task Learning with SE and Selective Weights,MTL_SSW),来联合学习语义分割和深度估计。MTL_SPP架构由共享骨干特征网络和任务特定的子网络组成,利用SE模块构建任务特定子网络,并利用金字塔池化增强特征提取。MTL_SSW在MTL_SPP的基础上,让任务特定子网络的语义分割特征和深度估计特征通过SW模块进行相互指导和优化,学习对特定任务更具判别性的特征。实验结果表明,提出的两种方法在NYUD_v2和SUNRGBD两个数据集上获得了优于先进方法的效果。展开更多
During a sea firing training,the intelligent detection of projectile-induced water column targets in a firing video is the prerequisite for and critical to the automatic calculation of miss distance,while the correct ...During a sea firing training,the intelligent detection of projectile-induced water column targets in a firing video is the prerequisite for and critical to the automatic calculation of miss distance,while the correct and precise calculation of miss distance is directly affected by the accuracy,false alarm rate and time delay of detection.After analyzing the characteristics of projectile-induced water columns,an accurate detection algorithm for time backtracked projectile-induced water columns based on the improved you only look once(YOLO)network is put forward.The capability and accuracy of detecting projectileinduced water column targets with the conventional YOLO network are improved by optimizing the anchor box through K-means clustering and embedding the squeeze and excitation(SE)attention module.The detection area is limited by adopting a sea-sky line detection algorithm based on gray level co-occurrence matrix(GLCM),so as to effectively eliminate such disturbances as ocean waves and ship wakes,and lower the false alarm rate of projectile-induced water column detection.The improved algorithm increases the mAP50 of water column detection by 30.3%.On the basis of correct detection,a time backtracking algorithm is designed with mean shift to track images containing projectile-induced water column in reverse time sequence.It accurately detects a projectile-induced water column at the time of its initial appearance as well as its pixel position in images,and considerably reduces detection delay,so as to provide the support for the automatic,accurate,and real-time calculation of miss distance.展开更多
In recent years,as intelligent transportation systems(ITS)such as autonomous driving and advanced driver-assistance systems have become more popular,there has been a rise in the need for different sources of traffic s...In recent years,as intelligent transportation systems(ITS)such as autonomous driving and advanced driver-assistance systems have become more popular,there has been a rise in the need for different sources of traffic situation data.The classification of the road surface type,also known as the RST,is among the most essential of these situational data and can be utilized across the entirety of the ITS domain.Recently,the benefits of deep learning(DL)approaches for sensor-based RST classification have been demonstrated by automatic feature extraction without manual methods.The ability to extract important features is vital in making RST classification more accurate.This work investigates the most recent advances in DL algorithms for sensor-based RST classification and explores appropriate feature extraction models.We used different convolutional neural networks to understand the functional architecture better;we constructed an enhanced DL model called SE-ResNet,which uses residual connections and squeeze-and-excitation mod-ules to improve the classification performance.Comparative experiments with a publicly available benchmark dataset,the passive vehicular sensors dataset,have shown that SE-ResNet outperforms other state-of-the-art models.The proposed model achieved the highest accuracy of 98.41%and the highest F1-score of 98.19%when classifying surfaces into segments of dirt,cobblestone,or asphalt roads.Moreover,the proposed model significantly outperforms DL networks(CNN,LSTM,and CNN-LSTM).The proposed RE-ResNet achieved the classification accuracies of asphalt roads at 98.98,cobblestone roads at 97.02,and dirt roads at 99.56%,respectively.展开更多
文摘微动脉瘤是糖尿病视网膜病变的初期症状,消除该病灶可在早期非常有效地预防糖尿病视网膜病变。但由于视网膜结构复杂,同时眼底图像的成像由于患者、环境、采集设备等因素的不同会存在不同的亮度和对比度,现有的微动脉瘤检测算法难以实现该病灶的精确检测和定位,为此本文提出嵌入SENet(squeeze-andexcitation networks)的改进YOLO(you only look once)v4自动检测算法。该算法在YOLOv4网络基础上,首先通过使用一种改进的快速模糊C均值聚类算法对目标样本进行先验框参数优化,以提高先验框与特征图的匹配度;然后,在主干网络嵌入SENet模块,通过强化关键信息,抑制背景信息,提高微动脉瘤的置信度;此外,还在网络颈部增加空间金字塔池化结构以增强主干网络输出特征的接受域,从而有助于分离出重要的上下文信息;最后,在Kaggle数据集上进行模型验证,并与其他方法进行对比。实验结果表明,与其他各种结构的YOLOv4网络模型相比,所提出的嵌入SENet的改进YOLOv4网络模型能显著提高检测结果(与原始YOLOv4相比Fscore提升了12.68%);与其他网络模型以及方法相比,所提出的嵌入SENet的改进YOLOv4网络模型的自动检测精度明显更优,且可实现精准定位。故本文所提出的嵌入SENet的改进YOLOv4算法性能较优,能准确、有效地检测并定位出眼底图像中的微动脉瘤。
文摘语义分割和深度估计任务是对图像像素级分类的研究,是两个高度相关的任务。从共享特征学习和特征交互融合两个角度出发,提出两个不同的多任务学习架构,即基于压缩激励模块(Squeeze-and-Excitation,SE)和金字塔池化的多任务学习网络(Multi-task Learning with SE and Pyramid Pooling,MTL_SPP),以及基于压缩激励和可选择权重(Selective Weight,SW)的多任务学习网络(Multi-task Learning with SE and Selective Weights,MTL_SSW),来联合学习语义分割和深度估计。MTL_SPP架构由共享骨干特征网络和任务特定的子网络组成,利用SE模块构建任务特定子网络,并利用金字塔池化增强特征提取。MTL_SSW在MTL_SPP的基础上,让任务特定子网络的语义分割特征和深度估计特征通过SW模块进行相互指导和优化,学习对特定任务更具判别性的特征。实验结果表明,提出的两种方法在NYUD_v2和SUNRGBD两个数据集上获得了优于先进方法的效果。
基金supported by the National Natural Science Foundation of China(51679247)。
文摘During a sea firing training,the intelligent detection of projectile-induced water column targets in a firing video is the prerequisite for and critical to the automatic calculation of miss distance,while the correct and precise calculation of miss distance is directly affected by the accuracy,false alarm rate and time delay of detection.After analyzing the characteristics of projectile-induced water columns,an accurate detection algorithm for time backtracked projectile-induced water columns based on the improved you only look once(YOLO)network is put forward.The capability and accuracy of detecting projectileinduced water column targets with the conventional YOLO network are improved by optimizing the anchor box through K-means clustering and embedding the squeeze and excitation(SE)attention module.The detection area is limited by adopting a sea-sky line detection algorithm based on gray level co-occurrence matrix(GLCM),so as to effectively eliminate such disturbances as ocean waves and ship wakes,and lower the false alarm rate of projectile-induced water column detection.The improved algorithm increases the mAP50 of water column detection by 30.3%.On the basis of correct detection,a time backtracking algorithm is designed with mean shift to track images containing projectile-induced water column in reverse time sequence.It accurately detects a projectile-induced water column at the time of its initial appearance as well as its pixel position in images,and considerably reduces detection delay,so as to provide the support for the automatic,accurate,and real-time calculation of miss distance.
基金funded by National Research Council of Thailand (NRCT):An Integrated Road Safety Innovations of Pedestrian Crossing for Mortality and Injuries Reduction Among All Groups of Road Users,Contract No.N33A650757supported by the Thailand Science Research and Innovation Fund+1 种基金the University of Phayao (Grant No.FF66-UoE001)King Mongkut’s University of Technology North Bangkok underContract No.KMUTNB-66-KNOW-05.
文摘In recent years,as intelligent transportation systems(ITS)such as autonomous driving and advanced driver-assistance systems have become more popular,there has been a rise in the need for different sources of traffic situation data.The classification of the road surface type,also known as the RST,is among the most essential of these situational data and can be utilized across the entirety of the ITS domain.Recently,the benefits of deep learning(DL)approaches for sensor-based RST classification have been demonstrated by automatic feature extraction without manual methods.The ability to extract important features is vital in making RST classification more accurate.This work investigates the most recent advances in DL algorithms for sensor-based RST classification and explores appropriate feature extraction models.We used different convolutional neural networks to understand the functional architecture better;we constructed an enhanced DL model called SE-ResNet,which uses residual connections and squeeze-and-excitation mod-ules to improve the classification performance.Comparative experiments with a publicly available benchmark dataset,the passive vehicular sensors dataset,have shown that SE-ResNet outperforms other state-of-the-art models.The proposed model achieved the highest accuracy of 98.41%and the highest F1-score of 98.19%when classifying surfaces into segments of dirt,cobblestone,or asphalt roads.Moreover,the proposed model significantly outperforms DL networks(CNN,LSTM,and CNN-LSTM).The proposed RE-ResNet achieved the classification accuracies of asphalt roads at 98.98,cobblestone roads at 97.02,and dirt roads at 99.56%,respectively.
