结合注意力的3D卷积网络脑胶质瘤分割算法

为了提升脑胶质瘤分割精度，提出一种结合注意力机制的3D卷积神经网络算法。输入3个不同尺度的图像块，经过9个卷积层和1个分类层后得到3个不同的分类结果，将分类结果与注意力学习到的权重相乘并逐体素相加得到输出。此外该算法采用了一种混合Dice损失函数与Focal损失函数的超参数损失函数。实验表明，该算法的Dice系数在整体区域、核心区域以及增强区域分别达到了95.31%、80.12%、82.25%。与已有的一种脑胶质瘤分割算法deepmedic相比，整体区域、核心区域以及增强区域的Dice系数分别提升了3%、2%、6%。在脑胶质瘤分割方面，具有重要的临床意义。     

基于卷积神经网络的单目深度估计

针对利用深度学习方法对街道图像进行深度估计，提出采用语义分割的方法解决深度图出现边界模糊等问题；估计深度通过左右视角图生成视差图进行无监督的训练。在网络模型中添加语义分割层，采取多个空洞卷积并行的结构增加感受野，同时减少了图像下采样的次数，降低了由于下采样带来的信息损失，使得的结果更加准确。这也是在深度估计中首次与空洞卷积相结合增加准确率。通过对KITTI街道数据集进行训练，与现有结果相比，除了增加检测准确性，降低错误率之外，使得效果图中的物体更加清晰，并且在效果图中还保留了一些原模型中被忽视掉的细节信息，将原始图像更加完整的表现出来。     

基于声源定位的高校在线巡课系统设计与研究

随着网络建设以及信息化教学方法在高校教学过程中的应用普及,越来越多的高校使用在线巡课系统对教师的教学过程进行跟踪和管理,以便发现课堂教学中的亮点、问题和不足。文章提出的在线巡课系统,基于声源定位的技术,对传统的在线巡课系统做出了改进,解决了已有巡课系统中“只闻其声,不见其人”的问题,能够更直观地跟踪到教师的教学过程,包括教学行为以及师生互动过程,有效提升教务人员巡课效果和体验感受。     

基于机器视觉的运煤车车号识别技术研究

在铁路运煤装车过程中为了快速、准确地识别车号,提出一种基于机器视觉的运煤车车号识别技术。将连通区域提取与投影分割法结合,实现车号的粗定位、细分割,并对图像中的断裂字符进行二次分割,构建了基于BP神经网络的分类模型进行车号识别,提升了煤炭装车的效率和精度。     

R2D2: A scalable deep learning toolkit for medical imaging segmentation

Deep learning has gained a significant popularity in recent years thanks to its tremendous success across a wide range of relevant fields of applications, including medical image analysis domain in particular. Although convolutional neural networks (CNNs) based medical applications have been providing powerful solutions and revolutionizing medicine, efficiently training of CNNs models is a tedious and challenging task. It is a computationally intensive process taking long time and rare system resources, which represents a significant hindrance to scientific research progress. In order to address this challenge, we propose in this article, R2D2, a scalable intuitive deep learning toolkit for medical imaging semantic segmentation. To the best of our knowledge, the present work is the first that aims to tackle this issue by offering a novel distributed versions of two well-known and widely used CNN segmentation architectures [ie, fully convolutional network (FCN) and U-Net]. We introduce the design and the core building blocks of R2D2. We further present and analyze its experimental evaluation results on two different concrete medical imaging segmentation use cases. R2D2 achieves up to 17.5× and 10.4× speedup than single-node based training of U-Net and FCN, respectively, with a negligible, though still unexpected segmentation accuracy loss. R2D2 offers not only an empirical evidence and investigates in-depth the latest published works but also it facilitates and significantly reduces the effort required by researchers to quickly prototype and easily discover cutting-edge CNN configurations and architectures.     

结合深度学习与支持向量机的金属零件识别

目的 在视觉引导的工业机器人自动拾取研究中，关键技术难点之一是机器人抓取目标区域的识别问题。特别是金属零件，其表面的反光、随意摆放时相互遮挡等非结构化因素都给抓取区域的识别带来巨大的挑战。因此，本文提出一种结合深度学习和支持向量机的抓取区域识别方法。方法 分别提取抓取区域的方向梯度直方图（HOG）和局部二进制模式（LBP）特征，利用主成分分析法（PCA）对融合后的特征进行降维，以此来训练支持向量机（SVM）分类器。通过训练Mask R-CNN（regions with convolutional neural network）神经网络完成抓取区域的初步分割。然后利用SVM对Mask R-CNN识别的抓取区域进行二次分类，完成对干扰区域的剔除。最后计算掩码完成实例分割，以此达到对抓取区域的精确识别。结果 对于随机摆放的铜质金属零件，本文算法与单一的Mask R-CNN及多特征融合的SVM算法就识别准确率、错检率、漏检率3个指标进行了比较，结果表明本文算法在识别准确率上较Mask R-CNN和SVM算法分别提高了7%和25%，同时有效降低了错检率与漏检率。结论 本文算法结合了Mask R-CNN与SVM两种方法，对于反光和遮挡情况具有一定的鲁棒性，同时有效地提升了目标识别的准确率。     

Automatic brain tissue segmentation in MR images using hybrid atlas forest based on confidence-weighted probability matrix

The segmentation of specific tissues in an MR brain image for quantitative analysis can assist the disease diagnosis and medical research. Therefore, a robust and accurate method for automatic segmentation is necessary. Atlas-based-method is a common and effective method of automatic segmentation where an atlas refers to a pair of image consist of an intensity image and its corresponding label image. Apart from the general multi-atlas-based methods, which propagate labels through the single atlas then fuse them, we proposed a hybrid atlas forest based on confidence-weighted probability matrix to consider the atlases set as a whole and treat each voxel differently. In the framework, we first register the atlas to the image space of target and calculate the confidence of voxels in the registered atlas. Then, a confidence-weighted probability matrix is generated and it augments to the intensity image of the atlas or target for providing spatial information of the target tissue. Third, a hybrid atlas forest is trained to gather the features and correlation information among the atlases in the dataset. Finally, the segmentation of the target tissues is predicted by the trained hybrid atlas forest. The segment performance and the components efficiency of the proposed method are evaluated on the two public datasets. Based on the experiment results and quantitative comparisons, our method can gather spatial information and correlation among the atlases to obtain an accurate segmentation.     

Hybrid Segmentation Scheme for Skin Features Extraction Using Dermoscopy Images

Objective and quantitative assessment of skin conditions is essential for cosmeceutical studies and research on skin aging and skin regeneration. Various handcraft-based image processing methods have been proposed to evaluate skin conditions objectively, but they have unavoidable disadvantages when used to analyze skin features accurately. This study proposes a hybrid segmentation scheme consisting of Deeplab v3+ with an Inception-ResNet-v2 backbone, LightGBM, and morphological processing (MP) to overcome the shortcomings of handcraft-based approaches. First, we apply Deeplab v3+ with an Inception-ResNet-v2 backbone for pixel segmentation of skin wrinkles and cells. Then, LightGBM and MP are used to enhance the pixel segmentation quality. Finally, we determine several skin features based on the results of wrinkle and cell segmentation. Our proposed segmentation scheme achieved a mean accuracy of 0.854, mean of intersection over union of 0.749, and mean boundary F1 score of 0.852, which achieved 1.1%, 6.7%, and 14.8% improvement over the panoptic-based semantic segmentation method, respectively.     

一种改进的最小二乘回归子空间分割方法

最小二乘回归(LSR)算法是一种常见的子空间分割方法，由于LSR具有解析解，因此它的聚类性能较高。然而LSR算法是应用谱聚类方法聚类数据，谱聚类方法初始化聚类中心是随机的，会影响后面的聚类效果。针对这一问题，提出一种基于聚类中心局部密度和距离这2个特点的改进的LSR算法（LSR-DC）。在Extended Yale B数据集上进行实验，结果表明，该算法有较高的聚类精度，具有一定的鲁棒性，优于现有LSR等子空间分割方法。     

基于GA-BP的汽车风振噪声声品质预测模型

目前对于汽车风振噪声的优化研究主要以声压级（Sound pressure level，SPL）作为单一评价指标，既不能全面反映噪声的物理属性，也无法考虑人耳对噪声的主观认知过程。为准确评价风振噪声，引入声品质，运用大涡模拟（Large eddy simulation，LES）对风振噪声进行数值仿真，根据实车道路试验判断仿真的准确性；对仿真结果进行声品质客观评价与主观评价，综合声品质客观评价参数与声品质主观评价试验结果建立BP神经网络预测模型；利用遗传算法（Genetic algorithm，GA），进一步对BP神经网络的结构参数进行优化，建立GA-BP声品质预测模型。研究结果表明，GA-BP声品质预测模型在训练速度和预测精度上都优于BP神经网络预测模型。预测模型基于声品质主客观评价结果，其预测值可以代替传统的声压级评价指标，为风振噪声提供更为准确合理的评价。