Convolutional neural network (CNN) is gradually applied to the field of image fusion because of its excellent performance. For the fusion task of infrared image and visible image, because there is no label data, unsupervised learning modeling is of great significance. To solve this problem, an unsupervised end-to-end depth fusion algorithm is proposed. The algorithm can directly predict the fused image containing significant information of the source image from the input infrared source image and visible light source image. The proposed algorithm constructs an auto-encoder network and uses real datasets for training. The loss function used in the network is the image structure similarity index measure (SSIM), which is widely used in image fusion tasks. Specifically, an improved non reference image evaluation index is designed to calculate the loss function, so as to achieve the purpose of unsupervised training of the network. In addition, the attention mechanism is introduced into the model to further improve the fusion results. The proposed algorithm is compared with many fusion algorithms, experimental results show that the fusion results of the proposed algorithm are very competitive in both subjective evaluation and objective index evaluation.
Abstract To solve the problem of trait differences in expression characteristics among different age groups, shortage of research on cross-age facial expression analysis, and the lack of datasets across ages, put forward the model in age factor deep sparse fusion extension of the expression recognition. The model utilizes a linear dictionary sequence of a high-level form as an input signal for a sparse representation, the input objects are constructed in a linear combination, and select the optimal solution in all solutions with the sparse level as an indicator, further through the convolutional, pooling and full connection processing of convolutional neural networks, the problems of high similarity, insufficient quantity and uneven distribution of data sets are solved through the fusion extension strategy, on this basis, the age operator is combined as feature elements and expression features are extracted and used as the basis for classification decision. Extension by deep sparse fusion, additional factors were extracted using anthropometric models, internal angle calculation and skin wrinkle detection of age operator as expression features, the comparison of results of multiple dataset experiments demonstrate the effectiveness and stability across datasets of the algorithm, the advantages of this algorithm are also demonstrated by the horizontal alignment of the representative cross-age expression classification algorithm. This method is effective for the accuracy and robustness of expression expressions across ages, it has the research value and the reference significance.
以美人蕉叶为原料,通过单因素试验并结合响应面法,对超声波辅助乙醇溶液提取美人蕉叶多酚的工艺条件进行研究.选择乙醇浓度、液料比、超声温度和超声时间作为响应面因素,多酚提取率为响应值,确定了能够用于美人蕉叶多酚提取率进行分析和预测的二次多项式回归模型方程.结果表明,在乙醇浓度为57%、液料比为43 mL/g、超声温度为72℃、超声时间为50 min时,测得的美人蕉叶多酚提取率最高,达到18.89 mg/g,与理论上的美人蕉叶多酚提取率(19.15 mg/g)相对误差仅为 1.36%,证实了该模型合理,预测可靠.所提取的美人蕉叶多酚对DPPH自由基和OH自由基具有一定的清除能力并呈现正相关关系,其IC50分别为 49.99 mg/L和 315.55 mg/L,证明了美人蕉叶多酚的抗氧化能力.
互联网技术的快速发展导致了互联网上数据信息的爆炸式增长.推荐系统作为解决互联网信息过载问题的关键技术,其核心思想是通过用户历史行为数据挖掘出用户的个性化偏好,为用户推荐其感兴趣的物品.然而,稀疏的评分数据会导致相似度计算不够准确,进而影响相似用户集的质量.为了提高相似用户搜索的可靠性,引入信任机制和评分子空间,提出基于评分子空间和信任机制的协同过滤推荐算法.创新点主要包括以下两点:首先,算法引入基于用户显式声明的关系数据所构建的信任机制,该关系数据能够对稀疏的评分数据进行补充.其次,利用评分子空间和信任关系,设计了一种基于隐式和显式相似度的混合相似度度量方式,并将之引入到多阶近邻的相似用户搜索方法和迭代评分预测方案中.实验结果表明,所提算法提高了推荐的准确度,具备较好的预测能力.
细粒度图像分类是计算机视觉领域的一大分类任务,其难点在于如何通过类别监督信息自主地找到判别性区域.提出一种新的通道-空间融合注意力模块,基于该模块设计了一种新的 Swin Transformer 算法SwinT-NCSA(a Swin Transformer based on a novel channel-spatial attention module),分别从通道维和空间维同时提取特征,再将其融入到Swin Transformer模型中以提高其小尺度中多头注意力信息的提取能力.SwinT-NCSA算法特别关注了对分类有用的区域,同时忽视对分类无用的背景区域,以此在细粒度图像分类任务中达到较高的分类准确率.在FGVC Aircraft飞机数据集、CUB-200-2011 鸟类数据集和Stanford Cars车类数据集3 个公共数据集上的实验表明,SwinT-NCSA算法可以分别取得 93.3%、88.4%和 94.7%的准确率,优于同类算法.
