Sentinel-2卫星的多光谱轻量级船舶目标检测算法

近年来深度卷积神经网络在可见光船舶检测方面取得了显著的进展，然而，大多数相关研究是通过改进大型的网络结构来提高检测性能，因此加大了对更高计算机性能的需求。此外，可见光图像难以在云、雾、海杂波、黑夜等复杂场景检测到船舶。针对以上问题，提出了一种融合红（red, R）、绿（green, G）、蓝（blue, B）和近红外（NIR）4个波段光谱信息的由粗到精细的轻量型船舶检测算法。与现有的方法中根据光谱特性利用水体检测算法提取水体区域不同之处是该算法是利用改进的水体检测算法来提取船舶候选区域。为获取更准确的候选区域，对船舶、厚云、薄云、平静海面、杂波海面5种场景中4个波段的像素值进行了统计分析，选取近红外大于阈值作为辅助判断，并以其中心点获取候选区域32×32大小的切片，并对切片进行非极大值抑制，由此获得了船舶粗检测结果。随后构建了轻量级LSGFNet网络对船舶候选区域切片进行精细识别。构建的网络融合了1×1卷积提取的波谱特征与3×3的提取几何特征，为防止光谱特征与几何特征的信息在融合时“信息不流通”，在LSGFNet网络中引入了ShuffleNet中的通道打乱机制，并减小了模型结构，与典型的轻量级网络相比具有更好的效果且模型较小。最后，利用Sentinel-2卫星多光谱10 m分辨率数据构建了512×512大小的1 120组数据进行粗检测，以及32×32大小的6 014组数据进行精细网络训练，其中候选区域粗提取的查全率为98.99%，精细识别网络精确度为96.04%，不同场景下的平均精确度为92.98%。实验表明该算法在抑制云层、海浪杂波等干扰的复杂背景下具有较高的检测效率，且训练时间短、计算机性能需求低。     

Subset representations and eigenvalues of the universal intertwining matrix

We solve a combinatorial question concerning eigenvalues of the universal intertwining endomorphism of a subset representation.     

On the domain of fractional Laplacians and related generators of Feller processes

In this paper we study the domain of the generator of stable processes, stable-like processes and more general pseudo- and integro-differential operators which naturally arise both in analysis and as infinitesimal generators of Lévy- and Lévy-type (Feller) processes. In particular we obtain conditions on the symbol of the operator ensuring that certain (variable order) Hölder and Hölder–Zygmund spaces are in the domain. We use tools from probability theory to investigate the small-time asymptotics of the generalized moments of a Lévy or Lévy-type process ${(X_t)}_{t\ge 0}$,

$\underset{t\to 0}{\lim }?\frac{1}{t}({\mathbb{E}}^{x}f(X_t)?f(x)),x\in {\mathbb{R}}^d,$

for functions f which are not necessarily bounded or differentiable. The pointwise limit exists for fixed $x\in {\mathbb{R}}^d$ if f satisfies a Hölder condition at x. Moreover, we give sufficient conditions which ensure that the limit exists uniformly in the space of continuous functions vanishing at infinity. As an application we prove that the domain of the generator of ${(X_t)}_{t\ge 0}$ contains certain Hölder spaces of variable order. Our results apply, in particular, to stable-like processes, relativistic stable-like processes, solutions of Lévy-driven SDEs and Lévy processes.     

Rate sensitivity and deformation mechanism of semi-crystalline polymers during instrumented indentation

Instrumented indentation tests using both constant loading rate (CLR) and continuous stiffness measurement (CSM) operation modes were performed to investigate the deformation mechanism and their sensitivity to the deformation rate in semi-crystalline polymers through the quantitative analysis of load-depth loading and unloading curves. The strain rate was constant during the CSM tests, while the strain rate decreased with the increasing of loading time in CLR tests. The mechanical response mechanism of the semi-crystalline polymers to these tests was very complicated because of the combined effects of strain-hardening in the crystal phase and strain-softening in the amorphous phase. Results show that the loading index m reflects the strain-hardening or strain-softening response during indentation. When m > 2, the mechanical response was due to the strain-hardening, and when m < 2, the response was due to strain-softening. A method based on the measured contact hardness was proposed to obtain the unloading stiffness, and the other mechanical parameters could then be determined according to the unloading stiffness.     

Modifying the physicochemical properties,solubility and foaming capacity of milk proteins by ultrasound-assisted alkaline pH-shifting treatment

This study investigated the effects of different treatment of alkaline pH-shifting on milk protein concentrate (MPC), micellar casein concentrate (MCC) and whey protein isolate (WPI) assisted by the same ultrasound conditions, including changes in the physicochemical properties, solubility and foaming capacity. The solubility of milk proteins had a significant increase with gradual enhancement of ultrasound-assisted alkaline pH-shifting (p < 0.05), especially for MCC up to 99.50 %. Also, treatment made a significant decline in the particle size of MPC and MCC, as well as the turbidity of the proteins (p < 0.05). The foaming capacity of MPC, MCC, and WPI was all improved, especially at pH 11, and at this pH, the milk protein also showed the highest surface hydrophobicity. The best foaming capacity at pH 11 was the result of the combined effect of particle size, potential, protein conformation, solubility, and surface hydrophobicity. In conclusion, ultrasound-assisted pH-shifting treatment was found to be effective in improving the physicochemical properties and solubility and foaming capacity of milk proteins, especially MCC, with promising application prospect in food industry.     

基于XGBOOST的恒星光谱分类特征数值化

恒星光谱分类是研究恒星的基础性工作之一，常用的光谱分类是基于20世纪70年代Morgan和Keenan建立起来的并逐步完善的MK分类系统。然而基于MK规则的交互式决策分类系统对处理海量天文光谱数据存在着一定的困难。目前光谱巡天一般采用的自动化分类则是模版匹配方法而忽略对谱线特征的测量。怎样自动、客观地提取海量光谱中的分类特征并应用这些特征进行分类可以对天体的物理化学性质的统计分析至关重要。针对此问题，通过机器学习和计算光谱的谱线指数结合的方法，提取光谱特征，并通过大数据分析定量地确定对光谱特征谱线的分类判据（数值化），确定每一类光谱具有物理意义的特征谱线的强度分布。首先对LAMOST DR4恒星光谱测量其谱线指数作为输入，光谱的分类标记采用官方发布的分类结果。使用XGBoost算法进行自动分类及特征排序，从而获得已知或未知的对于分类决策最为敏感的谱线。首先，选取高信噪比（S/N>30）、被LAMOST标记为B，A，F和M的恒星光谱数据，总计约414万个。然后，对光谱数据计算谱线指数从而使其得到降维处理，过滤冗余信息。其次，将处理后的恒星光谱数据随机划分为训练集和测试集，通过适当调整算法参数，用训练集得到所需要的分类决策树模型，用测试集测试其稳定性和可用性，以防止出现过拟合，同时使用算法自带函数进行提取分类特征。最后，输出并整理实验中算法所得的决策树模型，并挑选其概率比较大的分支作为最终的决策树模型。通过实验，可以发现在固定参数下，XGBoost所得的模型有一定的自适应性，较少受数据集影响，总体准确率可达88.5%；同时其所输出的分类决策树与已知的特征较为吻合，而且可以获得基于大数据的、数值化的特征谱线对应分类的范围，为完善基于特征的分类提供定量的规则。     

Selective Excitation of Band Extrema in Valley Photonic Crystals

Flexible control of building blocks of photonic crystals enables achieving desirable band structures. Exploration of photonic band extrema has brought many fantastic features to design artificial optical materials, such as Brillouin‐zone‐corner extrema for valley photonic materials and zone‐center extremum for zero‐index metamaterials. However, two such kinds of extrema are always found independently in different photonic crystals. In this work, a kind of valley photonic crystals possessing both zone‐center and zone‐corner band extrema almost at the same frequency is proposed. Inspired by antennas theory, a three‐antenna array (TAA) source is devoted to individually manipulate each extremum. The correlation coefficient is given to determine the coupling efficiency between the TAA source and extrema eigenmodes. By using a source with a high correlation coefficient, these extrema bulk states are selectively excited consistent with their eigenfields. Furthermore, three control cases are shown that multiple extrema points are simultaneously excited, in order to confirm the validity of the correlation coefficient. Finally, a potential application of a beam‐steering device is proposed through selective excitation of ternary extrema. This work develops binary valley states into ternary mix states, rendering more degrees of freedom for on‐chip optical information transport, particularly for beam steering and mode division multiplexing.