高光谱成像用高帧频CMOS图像传感器北大核心

针对高帧频、全局曝光和光谱平坦等成像应用需求,设计了一款高光谱成像用CMOS图像传感器。其光敏元采用PN型光电二极管,读出电路采用5T像素结构。采用列读出电路以及高速多通道模拟信号并行读出的设计方案来获得低像素固定图像噪声(FPN)和非均匀性抑制。芯片采用ASMC 0.35μm三层金属两层多晶硅标准CMOS工艺流片,为了抑制光电二极管的光谱干涉效应,后续进行了光谱平坦化VAE特殊工艺,并对器件的光电性能进行了测试评估。电路测试结果符合理论设计预期,成像效果良好,像素具备积分可调和全局快门功能,最终实现的像素规模为512×256,像元尺寸为30μm×30μm,最大满阱电子为400 ke^(-),FPN小于0.2%,动态范围为72 dB,帧频为450 f/s,相邻10 nm波段范围内量子效率相差小于10%,可满足高光谱成像系统对CMOS成像器件的要求。     

Dilated and soft attention-guided convolutional neural network for breast cancer histology images classification

Breast cancer is one of the most common types of cancer in women, and histopathological imaging is considered the gold standard for its diagnosis. However, the great complexity of histopathological images and the considerable workload make this work extremely time-consuming, and the results may be affected by the subjectivity of the pathologist. Therefore, the development of an accurate, automated method for analysis of histopathological images is critical to this field. In this article, we propose a deep learning method guided by the attention mechanism for fast and effective classification of haematoxylin and eosin-stained breast biopsy images. First, this method takes advantage of DenseNet and uses the feature map's information. Second, we introduce dilated convolution to produce a larger receptive field. Finally, spatial attention and channel attention are used to guide the extraction of the most useful visual features. With the use of fivefold cross-validation, the best model obtained an accuracy of 96.47% on the BACH2018 dataset. We also evaluated our method on other datasets, and the experimental results demonstrated that our model has reliable performance. This study indicates that our histopathological image classifier with a soft attention-guided deep learning model for breast cancer shows significantly better results than the latest methods. It has great potential as an effective tool for automatic evaluation of digital histopathological microscopic images for computer-aided diagnosis.     

As the number falls,alternatives to the Hagberg–Perten falling number method: A review

This review examines the application, limitations, and potential alternatives to the Hagberg–Perten falling number (FN) method used in the global wheat industry for detecting the risk of poor end-product quality mainly due to starch degradation by the enzyme α-amylase. By viscometry, the FN test indirectly detects the presence of α-amylase, the primary enzyme that digests starch. Elevated α-amylase results in low FN and damages wheat product quality resulting in cakes that fall, and sticky bread and noodles. Low FN can occur from preharvest sprouting (PHS) and late maturity α-amylase (LMA). Moist or rainy conditions before harvest cause PHS on the mother plant. Continuously cool or fluctuating temperatures during the grain filling stage cause LMA. Due to the expression of additional hydrolytic enzymes, PHS has a stronger negative impact than LMA. Wheat grain with low FN/high α-amylase results in serious losses for farmers, traders, millers, and bakers worldwide. Although blending of low FN grain with sound wheat may be used as a means of moving affected grain through the marketplace, care must be taken to avoid grain lots from falling below contract-specified FN. A large amount of sound wheat can be ruined if mixed with a small amount of sprouted wheat. The FN method is widely employed to detect α-amylase after harvest. However, it has several limitations, including sampling variability, high cost, labor intensiveness, the destructive nature of the test, and an inability to differentiate between LMA and PHS. Faster, cheaper, and more accurate alternatives could improve breeding for resistance to PHS and LMA and could preserve the value of wheat grain by avoiding inadvertent mixing of high- and low-FN grain by enabling testing at more stages of the value stream including at harvest, delivery, transport, storage, and milling. Alternatives to the FN method explored here include the Rapid Visco Analyzer, enzyme assays, immunoassays, near-infrared spectroscopy, and hyperspectral imaging.     

基于K-AP算法的高光谱图像波段选择方法

在高光谱图像分析领域中，波段选择是一种能有效减少高光谱图像维度的方法。K类仿射传播算法是一种高效的聚类算法，已成功地应用于人脸识别和数据分析等领域，但在高光谱图像分析领域还少有成功的应用。提出将K-AP算法应用于高光谱图像波段选择，对高光谱图像进行有效的数据压缩。针对K-AP算法的特点，基于Kullback-Leibler散度定义了新的相似度矩阵，对波段进行度量，再使用K-AP算法进行聚类，选择最有代表性的波段。实验结果表明，与常用的波段选择方法相比，所提出的方法有更好的表现。     

基于高光谱图像的改进SIFT特征提取与匹配

针对尺度不变特征变换(SIFT)算法所提取图像特征点数量少、误匹率高的问题,提出了一种基于高光谱图像的改进SIFT算法。首先,依据传统SIFT算法中高斯金字塔的构造思想,结合在不同波段下的高光谱图像具有相同宏观特征的特点,首次用高光谱图像作为原始算法中经高斯变换产生的图像,使得检测到的具有实际意义的特征点数量大幅增加;其次,传统SIFT算法以及大量的改进方法都只通过目标象元邻域范围内的像素信息来构造特征描述符,而忽略了像素点的位置信息,文中将目标象元的位置信息纳入了特征描述符,在特征描述符的匹配阶段,在利用邻域范围内的像素信息进行粗匹配之后,利用特征描述符中的位置信息进行精细匹配。仿真实验结果表明在限定最优值与次优值之比的情况下,采用高光谱图像构造高斯金字塔的方式能显著增加特征点的提取数量,更多地挖掘出图像中的极值点;在特征描述符中加入目标象元的位置信息作为特征点匹配第二阶段的判断依据,正确匹配数量达到原方法的59倍以上,极大提升了算法的匹配性能。