截面数据的B样条曲面重建研究

针对截面数据 ,研究其B样条曲面重建问题。给定节点矢量、依据最小二乘方法生成初始拟合曲面 ;为了提高拟合精度 ,采用迭代最近点方法优化修正数据点的参数 ,进而生成符合精度要求的拟合曲面。实际算例表明该算法简单、实用、可靠     

基于U-Net神经网络眼底图像的视盘分割研究

眼底图像中的视盘在青光眼筛查和诊断中起着重要作用。因此,从眼底图像中对视盘进行准确、快速地定位与分割具有重要意义。在过去,研究者们已经进行了对视盘的深入研究,但如何提高定位准确率和分割精度仍是视盘分割的一大难题。对此本文提出一种采用深度学习结构U-Net的视网膜视盘自动分割的方法,该方法结合机器学习,通过深度网络提取输入图像的视盘特征,从而得出相应的分割结果图。相对于传统的视盘分割方法,本文的U-Net神经网络能够有效学习有利于分割视盘的特征,从而提高分割的精确度,而且分割耗时更短。     

A method to measure the effective gas path length in the environmental or variable pressure scanning electron microscope

A simple method is described to determine the effective gas path length when incident electrons scatter in the gas above the specimen. This method is based on the measurement of a characteristic x-ray line emitted from a region close to the incident beam. From various experimental measurements performed on various microscopes, it is shown that the effective gas path length may increase with the chamber pressure and that it is also often dependent of the type of x-ray bullet.     

Two-phase flow in porous media: obtaining sharp digitized images of serial sections for subsequent quantitative analysis

Microscopic studies of the distributions of oil and aqueous phases within the pore space of reservoir rocks, at various relative saturations, are of fundamental importance in understanding migration, distribution, mobilization and recovery of oil in the context of a particular (enhanced) oil recovery method. The most effective approach to use for these studies is three-dimensional computer reconstruction with the data for the three-phase system (rock, oil, aqueous phase) entered by video digitization of a set of serial sections. A novel technique for creating accurate three-phase digitized images from the serial section data is described. By use of appropriate pore cast materials and lighting conditions, three different two-phase video images are created and digitized. In each image, a different one of the three phases in the pore cast is one of the two distinguishable phases. The two best two-phase digitized images are combined to give the three-phase digitized image.     

Three-dimensional reconstruction of microstructures in gibbsite using charge contrast images

An iterative process of serial sectioning and charge contrast imaging has been applied to gibbsite particles embedded in epoxy resin. The features observed in the two-dimensional sections have been reconstructed into three-dimensional (3-D) objects, so that the volumetric relationships in the microstructure are visualised. The 3-D objects confirm that the image detail present in charge contrast images of gibbsite is related to the microstructure and the processes occurring during the crystallisation of gibbsite. Speculative explanations for what these microstructures may represent are presented.     

Cryoultramicrotomy of muscle: improved preservation and resolution of muscle ultrastructure using negatively stained ultrathin cryosections

Ultrathin sections of rapidly frozen, briefly pre-treated muscle tissue are cut and thereafter are thawed and contrasted using a negative staining technique. The method has provided micrographs in which the in-vivo order in the muscle fibres has been preserved well enough to enable both a more complete interpretation of X-ray diffraction evidence from muscle, and also a gain of new ultrastructural information on aspects of myofibril and myofilament architecture in different types of fibre. Examples here are taken from chicken, rabbit and fish muscles and show both the M-band and the bridge region of the A-band in great detail. To enhance the detail in the original images, one-dimensional (1-D) and 2-D averaging techniques (lateral smearing and step averaging, respectively) are used. Although there is major shrinkage in section thickness to about one-third of its original value, demonstrated here for the first time is the fact that the characteristic A-band lattice planes are preserved in these sections in 3-D. This confirms the usefulness of cryosections not just for 1-D and 2-D image processing, but also for 3-D reconstruction. Thus, in combination with techniques of image processing, cryoultramicrotomy can give the muscle morphologist the detailed data that are needed to match the molecular biologists, biochemists and immunologists in the interpretation of their data about physiological and pathophysiological events in muscle fibres at the macromolecular level.     

改进Mask R-CNN的交通场景多目标快速检测与分割

针对智能驾驶中出现的交通场景多目标检测与分割效率低、鲁棒性差等问题,提出一种改进的Mask R-CNN交通场景多目标快速检测与分割方法。首先采用轻量级MobileNet作为骨干网络,有效减少网络参数并压缩模型体积,提升后续嵌入式端的算法移植能力,其次通过优化FPN与骨干网络卷积结构,保证高底层之间特征信息的完整传递,通过调整超参数得到交通场景多目标检测与分割改进网络模型。设计不同交通场景下的对比实验,改进网络能够准确实现多目标的检测与分割,平均检测精度可达85.2%。在ApolloScape和NuScence数据集上进行迁移实验,改进网络展示出良好的泛化能力。本文所提出的改进骨干网络与网络结构优化,能够适应多种复杂交通场景,完成交通场景多目标的快速检测与分割,为智能驾驶提供了理论依据与技术方案。     

基于DeepLabV3+与超像素优化的语义分割

针对基于深度学习的DeepLabV3+语义分割算法在编码特征提取阶段大量细节信息被丢失,导致其在物体边缘部分分割效果不佳的问题,本文提出了基于DeepLabV3+与超像素优化的语义分割算法。首先,使用DeepLabV3+模型提取图像语义特征并得到粗糙的语义分割结果;然后,使用SLIC超像素分割算法将输入图像分割成超像素图像;最后,融合高层抽象的语义特征和超像素的细节信息,得到边缘优化的语义分割结果。在PASCALVOC2O12数据集上的实验表明,相比较DeepLabV3+语义分割算法,本文算法在物体边缘等细节部分有着更好的语义分割性能,其mIoU值达到83.8%,性能得到显著提高并达到了目前领先的水平。     

Automatic measurement of volume percentage stroma in endometrial images using texture segmentation

The popularity of digital microscopy and tissue microarrays allow the use of high-throughput imaging for pathology research. To coordinate with this new technique, it is essential to automate the process of extracting information from such high amount of images. In this paper, we present a new model called the Subspace Mumford-Shah model for texture segmentation of microscopic endometrial images. The model incorporates subspace clustering techniques into a Mumford-Shah model to solve texture segmentation problems. The method first uses a supervised procedure to determine several optimal subspaces. These subspaces are then embedded into a Mumford-Shah objective function so that each segment of the optimal partition is homogeneous in its own subspace. The method outperforms a widely used method in bioimaging community called k-means segmentation since it can separate textures which are less separated in the full feature space, which confirm the usefulness of subspace clustering in texture segmentation. Experimental results also show that the proposed method is well performed on diagnosing premalignant endometrial disease and is very practical for segmenting image set sharing similar properties.     

Electron microscopy of semithick sections: Advantages for biomedical research

Many transmission electron microscopes are available which can be used to examine biological material in 0.25–0.50-μm-thick sections. When compared to the traditional thin section, these “semithick” sections possess a number of inherent advantages: They can be screened for content with the phase contrast light microscope, they facilitate many types of studies requiring an analysis of serial sections, and they are frequently the optimum thickness for stereomicroscopy. Structures such as microtubule-associated components, as well as structural relationships between cellular constituents, may also be clearly visible in semithick sections which are not visible, or go unnoticed, in thin sections. Together these advantages enable an investigator to obtain a more complete three-dimensional picture of a cell or cell component in a significantly (i.e., up to 90%) shorter period of time than would be required if thin sections were used. Semithick sections may, therefore, make a study feasible which is not approachable, or which is approachable only with great difficulty, by conventional thin sectioning techniques.     

基于高倍率细胞内镜系统的细胞核分割

临床上诊断消化道早期癌症主要依赖于电子内镜活检术,但是其诊断周期长。细胞内镜是一种具有超高放大倍率的内窥镜,配合术中染色可以直接在体内观察到病灶的细胞核等病理结构。为了使内窥镜医生能够在术中更准确地分析细胞核病理特征,基于已研制的高倍率细胞内镜系统在猪食管黏膜组织上开展了细胞核染色及分割方法研究。利用1%浓度的甲苯胺蓝水溶液对猪食管黏膜进行细胞核染色,并成功在细胞内镜显微成像模式下观察到染色的细胞核。在此基础上,采用深度学习方法训练了细胞核分割模型,有效实现了染色细胞核的分割提取,分割准确度达到了99.23%,特异性达到了99.54%,敏感性达到了84.37%,Dice系数达到了0.8138,为细胞内镜的AI辅助诊断算法研究奠定了基础。     

Dynamic Distribution Adaptation Based Transfer Network for Cross Domain Bearing Fault Diagnosis

In machinery fault diagnosis,labeled data are always difficult or even impossible to obtain.Transfer learning can lever-age related fault diagnosis knowledge from fully labeled source domain to enhance the fault diagnosis performance in sparsely labeled or unlabeled target domain,which has been widely used for cross domain fault diagnosis.How-ever,existing methods focus on either marginal distribution adaptation(MDA)or conditional distribution adaptation(CDA).In practice,marginal and conditional distributions discrepancies both have significant but different influences on the domain divergence.In this paper,a dynamic distribution adaptation based transfer network(DDATN)is pro-posed for cross domain bearing fault diagnosis.DDATN utilizes the proposed instance-weighted dynamic maximum mean discrepancy(IDMMD)for dynamic distribution adaptation(DDA),which can dynamically estimate the influ-ences of marginal and conditional distribution and adapt target domain with source domain.The experimental evalu-ation on cross domain bearing fault diagnosis demonstrates that DDATN can outperformance the state-of-the-art cross domain fault diagnosis methods.     

气溶胶粒子在矩形风管中沉降速度的试验研究

用滑石粉来模拟气溶胶粒子,对矩形风管系统(材料为镀锌钢板)的直管段和弯管段在不同的风速和粒径下,测定了不同风管表面的沉降速度,分析了它们之间的关系.试验结果表明,风管断面的沉降速度随风速和粒径的增大而增大;对于水平直管段,同一断面下底面的沉降速度最大,侧壁面次之,顶面最小;对于弯管段,外侧壁面和顶面的沉降速度比直管段有明显增加,内侧壁面减小.     

3D multiscale segmentation and morphological analysis of x‐ray microtomography from cold‐sprayed coatings

X‐ray microtomography from cold‐sprayed coatings brings a new insight on this deposition process. A noise‐tolerant segmentation algorithm is introduced, based on the combination of two segmentations: a deterministic multiscale segmentation and a stochastic segmentation. The stochastic approach uses random Poisson lines as markers. Results on a X‐ray microtomographic image of aluminium particles are presented and validated.