基于改进残差特征蒸馏的轻量级超分辨率网络

基于深度学习的图像超分辨率算法通常采用递归的方式或参数共享的策略来减少网络参数，这将增加网络的深度，使得运行网络花费大量的时间，从而很难将模型部署到现实生活中。为了解决上述问题，本文设计一种轻量级超分辨率网络，对中间特征的关联性及重要性进行学习，且在重建部分结合高分辨率图像的特征信息。首先，引入层间注意力模块，通过考虑层与层之间的相关性，自适应地分配重要层次特征的权重。其次，使用增强重建模块提取高分辨率图像中更精细的特征信息，以此得到更加清晰的重建图片。通过大量的对比实验表明，本文设计的网络与其他轻量级模型相比，有更小的网络参数量，并且在重建精度和视觉效果上都有一定的提升。     

超高效液相色谱⁃四极杆⁃飞行时间质谱对含PET食品接触材料中可迁移非挥发性物质的筛查研究

采用超高效液相色谱⁃四极杆⁃飞行时间高分辨质谱（UPLC⁃Q⁃TOF）对4类不同类型的含聚对苯二甲酸乙二醇酯（PET）材质的食品接触材料在4 %乙酸和50 %乙醇模拟物中的迁移出的非挥发性未知物进行筛查解析。结果表明，产品在4 %乙酸模拟物的迁移风险远小于50 %乙醇模拟物，主要迁移物质为聚合单体形成的寡聚物，抗氧剂、润滑剂、胶黏剂等加工助剂以及生产加工、迁移过程中形成的非有意添加物（NIAS）物质；纯PET材质的产品迁移物质较少，多层复合材料迁移物质较多。复合材质的产品中，PET材质可能在生产时添加了己二酸、癸二酸、新戊二醇等物质，进行了改性处理；此外，部分迁移物质会与模拟物中的乙醇发生反应，生成新的NIAS物质。     

Lactones in the Synthesis of Prostaglandins and Prostaglandin Analogs

In the total stereo-controlled synthesis of natural prostaglandins (PGs) and their structural analogs, a vast class of compounds and drugs, known as the lactones, are encountered in a few key steps to build the final molecule, as: δ-lactones, γ-lactones, and 1,9-, 1,11-, and 1,15-macrolactones. After the synthesis of 1,9-PGF_2α and 1,15-PGF_2α lactones, many 1,15-lactones of E₂, E₃, F₂, F₃, A₂, and A₃ were found in the marine mollusc Tethys fimbria and the quest for understanding their biological role stimulated the research on their synthesis. Then 1,9-, 1,11-, and 1,15-PG lactones of the drugs were synthesized as an alternative to the corresponding esters, and the first part of the paper describes the methods used for their synthesis. The efficient Corey procedure for the synthesis of prostaglandins uses the key δ-lactone and γ-lactone intermediates with three or four stereocenters on the cyclopentane fragment to link the PG side chains. The paper describes the most used procedures for the synthesis of the milestone δ-Corey-lactones and γ-Corey-lactones, their improvements, and some new promising methods, such as interesting, new stereo-controlled and catalyzed enantioselective reactions, and methods based on the chemical/enzymatic resolution of the compounds in different steps of the sequences. The many uses of δ-lactones not only for the synthesis of γ-lactones, but also for obtaining 9β-halogen-PGs and halogen-substituted cyclopentane intermediates, as synthons for new 9β-PG analogs and future applications, are also discussed.     

Surfactant-free electrochemical synthesis of fluoridated hydroxyapatite nanorods for biomedical applications

Fluoridated hydroxyapatite (FHA) [Ca₁₀(PO₄)₆F_x(OH)_2−x, x = 0–2] is believed to be a promising calcium phosphate (CaP) to replace pure hydroxyapatite (HA) for next-generation implants, owing to its better biocompatibility, higher antibacterial activity, and lower solubility. Notably, the shape and size of the CaP crystals play key roles in their performance and can influence their applications. One-dimensional (1D) FHA nanorods are important CaP materials which have been widely used in regenerative medicine applications such as restorative dentistry. Unfortunately, the traditional synthesis methods for FHA nanorods either employ surfactants or take a relatively long time. In this study, we aimed to propose a facile synthesis route to fabricate FHA nanorods without any surfactants using an electrochemical deposition method for the first time. This study focused on preparing FHA nanorods without the assistance of any surfactant, unlike the traditional synthesis methods, to avoid chemical impurities. FHA nanorods with lengths of 124–2606 nm, diameters of 28–211 nm, and aspect ratios of 4.4–21.8 were synthesized using the electrochemical method, followed by a heat treatment. For the as-synthesized FHA nanorods, the Ca/P ratio was 1.60 and the atomic concentration of F was 2.06 at.%. An ultrastructure examination revealed that each FHA nanorod possessed long-range order, good crystallinity, and a defect-free lattice with a certain crystallographic plane orientation along the whole rod. In short, we propose a novel, surfactant-free, cost-saving, and more efficient route to synthesize FHA nanorods which can be widely applied in multiple biomedical applications, including drug delivery, bone repair, and restorative dentistry.     

Production of (S)-β-Nitro Alcohols by Enantioselective C−C Bond Cleavage with an R-Selective Hydroxynitrile Lyase

Hydroxynitrile lyase (HNL)-catalysed stereoselective synthesis of β-nitro alcohols from aldehydes and nitroalkanes is considered an efficient biocatalytic approach. However, only one S-selective HNL—Hevea brasiliensis (HbHNL)—exists that is appropriate for the synthesis of (S)-β-nitro alcohols from the corresponding aldehydes. Further, synthesis catalysed by HbHNL is limited by low specific activity and moderate yields. We have prepared a number of (S)-β-nitro alcohols, by kinetic resolution with the aid of an R-selective HNL from Arabidopsis thaliana (AtHNL). Optimization of the reaction conditions for AtHNL-catalysed stereoselective C−C bond cleavage of racemic 2-nitro-1-phenylethanol (NPE) produced (S)-NPE (together with benzaldehyde and nitromethane, largely from the R enantiomer) in up to 99 % ee and with 47 % conversion. This is the fastest HNL-catalysed route known so far for the synthesis of a series of (S)-β-nitro alcohols. This approach widens the application of AtHNL for the synthesis not only of (R)- but also of (S)-β-nitro alcohols from the appropriate substrates. Without the need for the discovery of a new enzyme, but rather by use of a retro-Henry approach, it was used to generate a number of (S)-β-nitro alcohols by taking advantage of the substrate selectivity of AtHNL.     

基于GF-1数据的三峡库区水体信息精细化提取

为实现江河类狭长型水体信息的精细化提取，利用GF-1卫星数据，采用支持向量机和目视解译相结合的方法对三峡库区及重庆市水体信息进行了精细化提取。使用总体分类精度、Kappa系数、错分误差、漏分误差、制图精度和用户精度等指标对库区水体信息粗提取结果进行验证分析。结果表明:4个试验区水体提取的总体分类精度均超过90%，除试验区4的Kappa系数为0.884 1以外，其余试验区均超过0.9，提取精度较高。结合目视解译的方法，在粗提取结果的基础上对各问题进行精细化处理，得到精度高、完整性好的三峡库区以及重庆市水体信息数据，可为后续该区域的精细化遥感业务开展提供有效资料。     

Microtopography‐Guided Radial Gradient Circle Array Film with Nanoscale Resolution

Distribution of multimaterials at arbitrary positions with nanoscale resolution and over a large area substrate is essential to future advances in functional graded materials. Such stringent requirements are highly beyond the reach of current techniques, although newly developed 3D printing technologies are addressed. Here, a radial gradient circle array film with the distribution accuracy up to ≈18 nm is fabricated by using microtopographic substrate. A mathematical model is developed to guide the distribution of position, size, shape, and type of materials on an arbitrary section for the given morphology of substrate. The periodic electrical and mechanical properties of the radial gradient circle film are identified, which can be beneficial for further functionalization and applications, such as gradient refractive index lenses, microcoils, and microantennas.     

Preparation and characterization of multilayer Gd2O2S:Tb phosphor screen for X-ray detection application

Gd2O2S:Tb (GOS) phosphor screen, generally with a multilayer structure, has been widely used in digital X-ray imaging. GOS phosphor screen is fabricated via a slip casting route, which is an effective method to obtain multilayer composites with good strength. The GOS slurry was prepared using the menhaden oil, polyvinyl butyral and PEG-400 as dispersant, binder and plasticizer, respectively, and slip casted on a support layer. After the GOS scintillation layer is formed, a transparent polyethylene terephthalate film is covered on it as a protective layer. The X-ray excited emission spectra of the obtained GOS phosphor screen is well match to the spectral response of a-Si:H photodiode, and the spatial resolution was close to 3.7 LP/mm, which could meet the request of the commercial application.     

Robust image retrieval by cascading a deep quality assessment network

The performance of computer vision algorithms can severely degrade in the presence of a variety of distortions. While image enhancement algorithms have evolved to optimize image quality as measured according to human visual perception, their relevance in maximizing the success of computer vision algorithms operating on the enhanced image has been much less investigated. We consider the problem of image enhancement to combat Gaussian noise and low resolution with respect to the specific application of image retrieval from a dataset. We define the notion of image quality as determined by the success of image retrieval and design a deep convolutional neural network (CNN) to predict this quality. This network is then cascaded with a deep CNN designed for image denoising or super resolution, allowing for optimization of the enhancement CNN to maximize retrieval performance. This framework allows us to couple enhancement to the retrieval problem. We also consider the problem of adapting image features for robust retrieval performance in the presence of distortions. We show through experiments on distorted images of the Oxford and Paris buildings datasets that our algorithms yield improved mean average precision when compared to using enhancement methods that are oblivious to the task of image retrieval. ¹     

基于自回归谱外推的全聚焦成像分辨力提升

针对相邻缺陷全聚焦超声成像混叠问题，结合低阶、宽有效频带自回归谱外推方法，压缩超声波时域脉冲宽度，实现亚波长级全聚焦（Total focusing method，TFM）成像分辨力。建立碳钢试块模型，设置两个中心间距1.8 mm，直径1.3 mm圆孔，选用中心频率2.25 MHz，32阵元相控阵探头采集全矩阵数据。针对全矩阵数据，选择自回归阶数为2，信号频谱最大幅值下降14 dB为有效频带，建立自回归模型并外推有效频带外的高频与低频成分，随后对全矩阵数据进行延迟叠加处理和TFM成像。仿真结果表明，低阶、宽有效频带自回归谱外推处理方法具有较高的鲁棒性和准确性，TFM成像后可有效分离中心间距0.7λ（λ为超声波长）圆孔，保留缺陷横向位置信息的同时，定位误差不超过0.73%。对碳钢试块中相同位置及尺寸的圆孔进行试验验证，定位误差不超过1.06%，有效地提高TFM成像分辨力。