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。     

Growth of $${\text{N}}{{{\text{d}}}_{{{\text{1}}\; - \;y}}}{\text{Eu}}_{y}^{{{\text{2}} + }}{{{\text{F}}}_{{{\text{3}}\; - \;y}}}$$ Single Crystals with Tysonite-Type (LaF3) Structure and Investigation of the Concentration Dependence of Some Their Properties

Crystallography Reports - Nd1 – yEuyF3 – y crystals (y is the molar fraction of EuF2; 0 ≤ y ≤ 0.15) with a tysonite (LaF3) structure have...     

Synthesis of Substituted Aminopyrimidines as Novel Promising Tyrosine Kinase Inhibitors

Russian Journal of Organic Chemistry - A procedure has been proposed for the synthesis of a series of substituted N-(1,3-thiazol-2-yl)-pyrimidin-2-amines and N-(pyrimidin-2-yl)thioureas by...     

Formation and pharmacological activity of silicon—chitosan-containing glycerohydrogels obtained by biomimetic mineralization

The biomimetic sol—gel synthesis of silicon—chitosan-containing glycerohydrogels was carried out using silicon tetraglycerolate as a precursor. It was found that chitosan accelerates gel formation in weakly acidic media. In more acidic media, the kinetics of the process changes according to the curve with a maximum, which can be attributed to different mechanisms of silanol condensation before and after the isoelectric point. The investigated silicon—chitosan-containing glycerohydrogels exhibit antibacterial, anti-inflammatory, and wound healing activity. The synthesized hybrid glycerohydrogels are promising materials for biomedical applications.