Dual-Path Vision Transformer用于急性缺血性脑卒中辅助诊断

急性缺血性脑卒中是由于脑组织血液供应障碍导致的脑功能障碍,数字减影脑血管造影(DSA)是诊断脑血管疾病的金标准。基于患者的正面和侧面DSA图像,对急性缺血性脑卒中的治疗效果进行分级评估,构建基于Vision Transformer的双路径图像分类智能模型DPVF。为了提高辅助诊断速度,基于EdgeViT的轻量化设计思想进行了模型的构建;为了使模型保持轻量化的同时具有较高的精度,提出空间-通道自注意力模块,促进Transformer模型捕获更全面的特征信息,提高模型的表达能力;此外,对于DPVF的两分支的特征融合,构建交叉注意力模块对两分支输出进行交叉融合,促使模型提取更丰富的特征,从而提高模型表现。实验结果显示DPVF在测试集上的准确率达98.5%,满足实际需求。

Feature-Based Fusion of Dual Band Infrared Image Using Multiple Pulse Coupled Neural Network

To improve the quality of the infrared image and enhance the information of the object,a dual band infrared image fusion method based on feature extraction and a novel multiple pulse coupled neural network(multi-PCNN)is proposed.In this multi-PCNN fusion scheme,the auxiliary PCNN which captures the characteristics of feature image extracting from the infrared image is used to modulate the main PCNN,whose input could be original infrared image.Meanwhile,to make the PCNN fusion effect consistent with the human vision system,Laplacian energy is adopted to obtain the value of adaptive linking strength in PCNN.After that,the original dual band infrared images are reconstructed by using a weight fusion rule with the fire mapping images generated by the main PCNNs to obtain the fused image.Compared to wavelet transforms,Laplacian pyramids and traditional multi-PCNNs,fusion images based on our method have more information,rich details and clear edges.

An Efficient Combinatorial-Probabilistic Dual-Fusion Modification of Bernstein’s Polynomial Approximation Operator

The celebrated Weierstrass Approximation Theorem (1885) heralded intermittent interest in polynomial approximation, which continues unabated even as of today. The great Russian mathematician Bernstein, in 1912, not only provided an interesting proof of the Weierstrass’ theorem, but also displayed a sequence of the polynomials which approximate the given function . An efficient ‘Combinatorial-Probabilistic Dual-Fusion’ version of the modification of Bernstein’s Polynomial Operator is proposed. The potential of the aforesaid improvement is tried to be brought forth and illustrated through an empirical study, for which the function is assumed to be known in the sense of simulation.

Application of Dual-Energy CT Non-Linear Fusion Technology in Improving CTA Image Quality of Renal Cancer

Objective: To explore the significance of dual-energy CT non-linear fusion technique in improving the quality of CTA image of renal cancer. Methods: The CTA images of 100 patients who had been confirmed by pathology as renal cancer were collected and were randomly divided into experimental group and control group with 50 cases respectively. The two groups of patients were treated with iodine concentration of 300 mg/ml and 350 mg/ml non-ionic contrast agent, with a dosage of 1.5 ml/kg and an injection rate of 4 ml/s. The contrast agent intelligently tracking method was adopted bolus. The control group used the conventional CTA scanning, with a reference tube voltage/tube current of 100 kv/ref150 mas. The experimental group adopted the double energy scanning, with ball tube A and ball tube B. The reference tube voltage/tube current was 100 kv/ref250 mas and sn150 kv/ref125 mas respectively. The images of the experimental group were non-linear fused to obtain the Mono+ 55 kev single-energy images. The CT value, SNR contrast ratio of the abdominal aorta, renal artery and tumor tissue of the experimental group images and the 100 KV images and the Mono+ 55 kev images of the control group were compared. The objective evaluation and subjective evaluation of the image quality of the three groups of images was performed. Results: The results showed that the 100 kV images of the experimental group were statistically different from those of the control group (P05) in CT value, SNR and CNR (P 0.05). And there was no statistically significant difference between the non-linear fusion single-energy Mono+ 55 kev images and the control group images in CT value, SNR and CNR (P > 0.05). The subjective evaluation of image quality showed that there was no significant difference between Mono+ 55 kev images and control group images, and the quality of Mono+ 55 kev images was higher than that of experimental group 100 kV images. Conclusion: The dual-energy CT non-linear fusion technique can improve the quality of CTA image in patients with renal cancer, and it is possible to obtain high quality CTA images with low iodine concentration contrast agent.

Neutronics Optimization of LiPb-He Dual-Cooled Fuel Breeding Blanket for the Fusion-Driven sub-critical System

The concept of the liquid Li17Pb83 and Helium gas dual-cooled Fuel Breeding Blanket (FBB) for the Fusion-Driven sub-critical System (FDS) is presented and analyzed. Taking self-sustaining tritium (TBR >1.05) and annual output of 100 kg or more fissile 239Pu (FBR > 0.238) as objective parameters, and based on the three-dimensional Monte Carlo neutron-photon transport code MCNP/4A, a neutronics-optimizated calculation of different cases was carried out and the concept is proved feasible. In addition, the total breeding ratio ( BR = TBR + FBR ) is listed corresponding to different cases.

A Progressive Feature Fusion-Based Manhole Cover Defect Recognition Method

Manhole cover defect recognition is of significant practical importance as it can accurately identify damaged or missing covers, enabling timely replacement and maintenance. Traditional manhole cover detection techniques primarily focus on detecting the presence of covers rather than classifying the types of defects. However, manhole cover defects exhibit small inter-class feature differences and large intra-class feature variations, which makes their recognition challenging. To improve the classification of manhole cover defect types, we propose a Progressive Dual-Branch Feature Fusion Network (PDBFFN). The baseline backbone network adopts a multi-stage hierarchical architecture design using Res-Net50 as the visual feature extractor, from which both local and global information is obtained. Additionally, a Feature Enhancement Module (FEM) and a Fusion Module (FM) are introduced to enhance the network’s ability to learn critical features. Experimental results demonstrate that our model achieves a classification accuracy of 82.6% on a manhole cover defect dataset, outperforming several state-of-the-art fine-grained image classification models.

ON THE STABILITY OF FUSION FRAMES (FRAMES OF SUBSPACES)

A frame is an orthonormal basis-like collection of vectors in a Hilbert space, but need not be a basis or orthonormal. A fusion frame （frame of subspaces） is a frame-like collection of subspaces in a Hilbert space, thereby constructing a frame for the whole space by joining sequences of frames for subspaces. Moreover the notion of fusion frames provide a framework for applications and providing efficient and robust information processing algorithms.In this paper we study the conditions under which removing an element from a fusion frame, again we obtain another fusion frame. We give another proof of [5, Corollary 3.3（iii）] with extra information about the bounds.

结合视觉显著性与Dual-PCNN的红外与可见光图像融合

针对现存的红外与可见光图像融合算法亮度不均、目标不突出、对比度不高、细节丢失等问题,结合非下采样剪切波变换(NSST)具有多尺度、最具稀疏表达的特性,显著性检测具有突出红外目标的优势,双通道脉冲耦合神经网络(Dual-PCNN)具有耦合、脉冲同步激发等优点,提出一种基于NSST结合视觉显著性引导Dual-PCNN的图像融合方法。首先,通过NSST分解红外与可见光图像各方向的高频与低频子带系数;然后,低频子带系数采用基于显著性决策图引导Dual-PCNN融合策略,高频子带系数采用改进的空间频率作为优化Dual-PCNN的激励进行融合;最后,经过NSST逆变换得到融合图像。实验结果表明,融合图像红外目标突出且可见光背景细节丰富。该方法相比于其他融合算法在主观评价与客观评价上都有一定程度的改善。

Fusion Platform for Renal Cyst Characterization

This work concerns the field of diagnostic aids that facilitate diagnostic decisions for practitioners, especially in medical imaging. The pathology in question, in this study, is the renal cyst. The diagnostic process starts from simultaneous acquisitions of double isotope (Teechnetium-99 m and Iodine-131) scintigraphic images. Then, the platform allows the fusion of these images and the calculation of a pathological parameter that permits the characterization of the state of the dysplasic kidney by comparing it with the normal one. The final result is fusion images annotated by the pathological parameter value.

Multiscale Based Local Structurization Information Metric for Robust Pixel Level Image Fusion

Because previous methods can not identify underlying image features from noises effectively, the updated image fusion schemes will be degraded when inputs are corrupted with noise. The perceptual salient image features often manifest some geometric structures, while noise dominated images are less structured. Based on complex wavelet transform, a structurization information metric is formulated by means of the Von Neumann entropy. The formulated metric can distinguish image features from noise very well. During the fusion process, the metric is employed to weight all fusion inputs. As a result, the perceptual meaningful inputs are enhanced while the noise inputs are de-emphasized adaptively. Comparing several image fusion schemes subjectively and objectively shows the good performance of the new scheme.

双通道特征融合的真实场景点云语义分割方法

真实场景点云不仅具有点云的空间几何信息,还具有三维物体的颜色信息,现有的网络无法有效利用真实场景的局部特征以及空间几何特征信息,因此提出了一种双通道特征融合的真实场景点云语义分割方法DCFNet(dual-channel feature fusion of real scene for point cloud semantic segmentation)可用于不同场景下的室内外场景语义分割。更具体地说,为了解决不能充分提取真实场景点云颜色信息的问题,该方法采用上下两个输入通道,通道均采用相同的特征提取网络结构,其中上通道的输入是完整RGB颜色和点云坐标信息,该通道主要关注于复杂物体对象场景特征,下通道仅输入点云坐标信息,该通道主要关注于点云的空间几何特征;在每个通道中为了更好地提取局部与全局信息,改善网络性能,引入了层间融合模块和Transformer通道特征扩充模块;同时,针对现有的三维点云语义分割方法缺乏关注局部特征与全局特征的联系,导致对复杂场景的分割效果不佳的问题,对上下两个通道所提取的特征通过DCFFS(dual-channel feature fusion segmentation)模块进行融合,并对真实场景进行语义分割。对室内复杂场景和大规模室内外场景点云分割基准进行了实验,实验结果表明,提出的DCFNet分割方法在S3DIS Area5室内场景数据集以及STPLS3D室外场景数据集上,平均交并比(MIOU)分别达到71.18%和48.87%,平均准确率(MACC)和整体准确率(OACC)分别达到77.01%与86.91%,实现了真实场景的高精度点云语义分割。

域迁移增强的综合假脸检测模型

提出一种跨域的综合假脸检测模型。首先,设计一种双域融合模型,该模型利用空间注意力机制实现RGB域和频域特征的融合。其次,在此基础上,结合数据增强技术,提出了一种跨域迁移策略。最后,提出的双域模型的精度,在5个通用数据集上均比单域模型有一定的提高,尤其在NT数据集上,该方法的精度比EfficientNet-B0方法提高了3.4%。此外,实验结果表明,与其他迁移学习方法相比,在FaceForensics++和Celeb-df数据集上,该方法在域迁移中具有更好的泛化性能。

多示例学习的簇频繁性分析及双角度融合嵌入

多示例学习(Multi-Instance Learning,MIL)的训练数据是由若干个未带标记的示例组成的带标记的包,基于嵌入的方法,通过将包嵌入成单向量来解决包表示问题,然而大部分现有方法忽略了示例与包的联系,难以保证所选示例的代表性.同时,单角度的嵌入方法无法有效地提取正、负包的差异信息,使嵌入向量的质量较差.提出一种多示例学习的簇频繁性分析及双角度融合嵌入(FADE).簇频繁性分析技术从正、负子空间中分别筛选部分示例作为子空间的簇心,依据簇心将子空间聚类成簇,再计算簇频繁性指标,选择频繁性较高的簇的簇心组成子空间代表示例集.双角度融合嵌入技术基于正、负子空间代表示例集和差值嵌入函数,分别从正、负角度挖掘信息,融合两个角度信息获得最终的嵌入向量.在29个数据集上与七个MIL算法进行了对比实验,结果表明,FADE的分类准确率总体上优于七个对比算法,在图像数据集上有显著优势,在文本和网页数据集上也表现良好.

Multi-Feature Fusion Based Relative Pose Adaptive Estimation for On-Orbit Servicing of Non-Cooperative Spacecraft

On-orbit servicing, such as spacecraft maintenance, on-orbit assembly, refueling, and de-orbiting, can reduce the cost of space missions, improve the performance of spacecraft, and extend its life span. The relative state between the servicing and target spacecraft is vital for on-orbit servicing missions, especially the final approaching stage. The major challenge of this stage is that the observed features of the target are incomplete or are constantly changing due to the short distance and limited Field of View (FOV) of camera. Different from cooperative spacecraft, non-cooperative target does not have artificial feature markers. Therefore, contour features, including triangle supports of solar array, docking ring, and corner points of the spacecraft body, are used as the measuring features. To overcome the drawback of FOV limitation and imaging ambiguity of the camera, a "selfie stick" structure and a self-calibration strategy were implemented, ensuring that part of the contour features could be observed precisely when the two spacecraft approached each other. The observed features were constantly changing as the relative distance shortened. It was difficult to build a unified measurement model for different types of features, including points, line segments, and circle. Therefore, dual quaternion was implemented to model the relative dynamics and measuring features. With the consideration of state uncertainty of the target, a fuzzy adaptive strong tracking filter( FASTF) combining fuzzy logic adaptive controller (FLAC) with strong tracking filter(STF) was designed to robustly estimate the relative states between the servicing spacecraft and the target. Finally, the effectiveness of the strategy was verified by mathematical simulation. The achievement of this research provides a theoretical and technical foundation for future on-orbit servicing missions.

拉普拉斯卷积的双路径特征融合遥感图像智能解译方法

由于遥感图像存在多尺度变化和目标边缘模糊等问题,对其进行智能解译仍然是一项极具挑战性的工作。传统的语义分割方法在处理这些问题时存在局限性,难以有效捕捉全局和局部信息。针对上述问题,文中提出一种双路径特征融合分割方法 DFNet。首先,使用Swin Transformer作为主干提取全局语义特征,以处理像素之间的长距离依赖关系,从而促进对图像中不同区域相关性的理解;其次,将拉普拉斯卷积嵌入到空间分支,以捕获局部细节信息,加强目标地物边缘信息表达;最后,引入多尺度双向特征融合模块,充分利用图像中的全局和局部信息,以增强多尺度信息的获取能力。在实验中,使用了三个公开的高分辨率遥感图像数据集进行验证,并通过消融实验验证了所提模型不同模块的作用。实验结果表明,所提方法在Uavid数据集、Potsdam数据集、LoveDA数据集的mIoU达到了71.32%、85.58%、54.01%,提高了语义分割的性能,使分割结果更为精细。

面向小样本抽取式问答的多标签语义校准方法

小样本抽取式问答任务旨在利用文章给定的上下文片段,抽取出真实的答案片段。其基线模型采用的方法只针对跨度进行学习,缺乏对全局语义信息的利用,在含有多组不同重复跨度的实例中存在着理解偏差等问题。为了解决上述问题,该文利用不同层级的语义提出了一种面向小样本抽取式问答任务的多标签语义校准方法。采用包含全局语义信息的头标签和基线模型中的特殊字符构成多标签进行语义融合,并利用语义融合门来控制全局信息流的引入,将全局语义信息融合到特殊字符的语义信息中。然后,利用语义筛选门对新融入的全局语义信息和该特殊字符的原有语义信息进行保留与更替,实现对标签偏差语义的校准。在8个小样本抽取式问答数据集中的56组实验结果表明:该方法在评价指标F1值上均明显优于基线模型,证明了所提方法的有效性和先进性。

采用级联策略融合边界特征的多尺度息肉分割网络

结直肠息肉分割能有效辅助医生筛查大肠腺瘤,但息肉分割存在噪声较多、边界区分度不够等问题。针对以上问题,本文设计了一种采用级联策略融合边界特征的多尺度息肉分割网络。首先,本文提出了一种改进的通道分组空间增强模块,以增强骨干网络提取的图像特征,从而提高通道和空间位置的相关性。其次,考虑到边界区分度不够,设计了一个级联特征融合网络,以更好地保留边界信息并提高边界区分度,从而提高分割精度。最后,引入了一种双分支混合上采样模块来获取更多的特征细节信息,以实现特征的互补以及捕获更完整有效的特征。在CVC-ClinicDB和Kvasir数据集上进行测试,本文算法的平均Dice系数分别为0.944,0.920,平均交并比分别为0.900,0.869;而M2SNet算法的平均Dice系数分别为0.922,0.912,平均交并比分别为0.880,0.861。在ETIS-LaribPolypDB,CVC-300和CVC-ColonDB数据集上进行测试,本文算法的平均Dice系数分别为0.776,0.915,0.782;而M2SNet算法的平均Dice系数分别为0.749,0.903,0.758。实验结果表明本文算法的分割精度较高,泛化能力较强。

基于光度立体和双流特征融合网络的工业产品表面缺陷检测方法

表面缺陷检测是现代工业生产流程中的重要环节。现有的视觉缺陷检测方法一般通过对目标对象的单幅RGB或灰度图像进行分析,利用缺陷与背景之间的差异性特征实现检测,适用于目标与背景呈较大区别的对象,如金属表面的氧化、斑点缺陷检测。但单纯的RGB图像无法有效地表征主要由深度变化形成的凹坑、凸包等3维缺陷特征,最终导致漏检。为此,文中根据多方向光照成像及光度立体原理提取待测对象表面的3维几何形貌信息;接着,利用对比度金字塔融合算法对原始的多方向光照图像进行有效融合,得到增强的缺陷的2维RGB融合图像特征;然后,在多目标检测框架YOLOv5的基础上,以上述几何形貌及RGB融合图像为输入,构建一种基于双流特征融合的缺陷检测网络模型,该模型引入了空间通道注意力残差模块和门控循环单元特征融合模块,能在多个层级对不同模态特征进行有机融合,实现对表面缺陷的2维RGB及3维形貌信息的有效提取,达到同时应对2维和3维缺陷检测的目的;最后对若干典型工业产品表面缺陷进行检测实验。结果表明,文中方法在多个数据集上的平均检测准确率均超过90%,且能同时应对2维、3维缺陷的检测,检测性能优于目前的主流方法,能够适应不同工业产品表面的检测需求。

基于细节增强的双分支实时语义分割网络

实时语义分割方法常利用双分支结构分别保存图像的浅层空间信息和深层语义信息。然而,当前基于双分支结构的实时语义分割方法重点研究语义特征的挖掘,忽略了空间特征的保持,导致网络无法精准地捕捉图像内物体的边界和纹理等细节特征,最终分割效果欠佳。针对以上问题,提出基于细节增强的双分支实时语义分割网络(DEDBNet),多阶段增强空间细节信息。首先,提出细节增强双向交互(DEBIM)模块,在分支间的交互阶段使用轻量空间注意力机制增强高分辨率特征图对细节信息的表达能力,促进空间细节特征在高低两分支上的流动,以加强网络对细节信息的学习能力;其次,设计局部细节注意力特征融合模块(LDAFF),在两分支末端特征融合的过程中同时建模全局语义信息和局部空间信息,解决不同层次特征图之间细节不连续的问题;此外,引入边界损失,在不影响模型速度的情况下引导网络浅层学习物体边界信息。所提网络在Cityscapes验证集上以92.3 frame/s的帧速率(FPS)获得78.2%的平均交并比(mIoU),在CamVid测试集上以202.8 frame/s获得79.2%的mIoU;与深度双分辨率网络(DDRNet-23-slim)相比,mIoU分别提高了1.1和4.5个百分点。实验结果表明,DEDBNet能够准确地分割场景图像,且满足实时性要求。

基于Elmo和注意力机制的双通道文本分类模型

针对中文文本分类过程中文本特征提取不全面、语义表征不准确的问题,提出一种基于改进Elmo模型、带有注意力机制的卷积神经网络与门控循环网络相结合的双通道文本分类模型。模型首先将静态词向量输入Elmo模型生成动态词向量对文本进行表示;然后利用双通道结构构建加入注意力机制的卷积神经网络和双向门控循环网络分别提取文本内部特征和全局语义信息;最后,将双通道特征向量融合处理后通过分类器完成文本分类。依托THUCNews数据集进行模型的仿真,所提模型分类准确率和召回率分别为90.21%、90.45%,实验结果表明,与其它分类模型相比,所提模型具有更好的分类性能。