Detection of Precipitation Cloud over the Tibet Based on the Improved U-Net

Aiming at the problem of radar base and ground observation stations on the Tibet is sparsely distributed and cannot achieve large-scale precipitation monitoring. UNet, an advanced machine learning (ML) method, is used to develop a robust and rapid algorithm for precipitating cloud detection based on the new-generation geostationary satellite of FengYun-4A (FY-4A). First, in this algorithm, the real-time multi-band infrared brightness temperature from FY-4A combined with the data of Digital Elevation Model (DEM) has been used as predictor variables for our model. Second, the efficiency of the feature was improved by changing the traditional convolution layer serial connection method of U-Net to residual mapping. Then, in order to solve the problem of the network that would produce semantic differences when directly concentrated with low-level and high-level features, we use dense skip pathways to reuse feature maps of different layers as inputs for concatenate neural networks feature layers from different depths. Finally, according to the characteristics of precipitation clouds, the pooling layer of U-Net was replaced by a convolution operation to realize the detection of small precipitation clouds. It was experimentally concluded that the Pixel Accuracy (PA) and Mean Intersection over Union (MIoU) of the improved U-Net on the test set could reach 0.916 and 0.928, the detection of precipitation clouds over Tibet were well actualized.     

基于3D和1D多特征融合的语音情感识别算法

针对语音情感识别任务中特征提取单一、分类准确率低等问题,提出一种3D和1D多特征融合的情感识别方法,对特征提取算法进行改进。在3D网络,综合考虑空间特征学习和时间依赖性构造,利用双线性卷积神经网络（Bilinear Convolutional Neural Network,BCNN）提取空间特征,长短期记忆网络（Short-Term Memory Network,LSTM）和注意力（attention）机制提取显著的时间依赖特征。为降低说话者差异的影响,计算语音的对数梅尔特征（Log-Mel）和一阶差分、二阶差分特征合成3D Log-Mel特征集。在1D网络,利用一维卷积和LSTM的框架。最后3D和1D多特征融合得到判别性强的情感特征,利用softmax函数进行情感分类。在IEMOCAP和EMO-DB数据库上实验,平均识别率分别为61.22%和85.69%,同时与提取单一特征的3D和1D算法相比,多特征融合算法具有更好的识别性能。     

基于无监督学习的高分辨光声重建

光声断层成像(Optoacoustic Tomography,OAT)是一种新兴的生物医学成像技术,在基础医学研究与临床实践中具有重要作用。针对现有光声断层成像空间分辨率较低的问题,提出了一种结合物理点扩散函数(Point Spread Function,PSF)模型和卷积神经网络(Convolutional Neural Network,CNN)的新型高分辨光声重建网络方法(Physical Attention U-Net,Phys-AU-Net)。该方法采用无监督学习策略,结合物理PSF模型和基于注意力机制的U-Net网络。其中,物理PSF模型用于完成对衍射受限机制的模拟,基于注意力机制的U-Net网络用于实现对高密度重叠吸收体图像的特征提取。在二者共同作用下,Phys-AU-Net突破了声衍射极限对于OAT成像空间分辨率的限制。实验结果表明,Phys-AU-Net能够有效实现对声衍射受限光声断层图像的高分辨重建,其性能相较于U-Net网络具有较大程度提升,在结构相似性指标(Structural Similarity,SSIM)方面提升了43.5%,在峰值信噪比(Peak Sign...     

基于环形特征的卷积神经网络轮毂识别

针对实际生产中不同种类轮毂的混流生产问题,提出了一种基于环形特征的卷积神经网络轮毂识别算法。将直角坐标下的环形轮毂映射到极坐标中,归一化为标准形式的矩形,提取轮毂图像的环形特征信息,减少冗余特征产生的影响;设计了一种改进的VGG网络架构,利用深度可分离卷积打破输出通道维度与卷积核大小的联系,在不损失网络性能的同时降低了计算量,能够在实际生产中轮毂识别任务在有限的算力情况下实时进行计算;从有效性和实时性两个方面对轮毂识别算法进行评估,且通过Inception V3、SVM、KNN等模型的对比实验,验证了该算法可以实时地对轮毂自适应分类。实验表明: 该方法对轮毂图像的处理精度达到99%以上,单幅图像平均处理时间降低至11.78ms。     

基于稀疏激光点云数据和单帧图像融合的三维重构算法

由于激光雷达获取的深度数据非常稀疏,为了能够将深度数据与图像数据重构出稠密三维深度图,本文提出了基于稀疏激光点云数据和单帧图像融合的三维重构算法。该方法首先使用点直方图特征有效地选择对应于目标的点数据并消除体素中的非相似点;然后,使用高斯过程回归对局部深度数据建模,并通过插值获得三维深度数据,本文算法获得的三维深度点更接近基准值,并保持了目标的局部形状特征;最后,利用马尔科夫随机场对图像灰度数据和三维插值点进行融合来构建三维深度图。仿真实验结果表明:相比现有基于激光雷达数据和单目图像数据的三维重建算法,本文提出的算法将大大提升算法的鲁棒性与重构的准确度,可辅助用于复杂的城市场景中车辆的无人驾驶。     

MFF-Net: Multimodal Feature Fusion Network for 3D Object Detection

In complex traffic environment scenarios, it is very important for autonomous vehicles to accurately perceive the dynamic information of other vehicles around the vehicle in advance. The accuracy of 3D object detection will be affected by problems such as illumination changes, object occlusion, and object detection distance. To this purpose, we face these challenges by proposing a multimodal feature fusion network for 3D object detection (MFF-Net). In this research, this paper first uses the spatial transformation projection algorithm to map the image features into the feature space, so that the image features are in the same spatial dimension when fused with the point cloud features. Then, feature channel weighting is performed using an adaptive expression augmentation fusion network to enhance important network features, suppress useless features, and increase the directionality of the network to features. Finally, this paper increases the probability of false detection and missed detection in the non-maximum suppression algorithm by increasing the one-dimensional threshold. So far, this paper has constructed a complete 3D target detection network based on multimodal feature fusion. The experimental results show that the proposed achieves an average accuracy of 82.60% on the Karlsruhe Institute of Technology and Toyota Technological Institute (KITTI) dataset, outperforming previous state-of-the-art multimodal fusion networks. In Easy, Moderate, and hard evaluation indicators, the accuracy rate of this paper reaches 90.96%, 81.46%, and 75.39%. This shows that the MFF-Net network has good performance in 3D object detection.     

Three dimension reconstruction through measure-based image selection

Three dimension (3D) reconstruction is one of the research focus of computer vision and widely applied in various fields. The main steps of 3D reconstruction include image acquisition, feature point extraction and matching, camera calibration and production of dense 3D scene models. Generally, not all the input images are useful for camera calibration because some images contain similar and redundant visual information. These images can even reduce the calibration accuracy. In this paper, we propose an effective image selection method to improve the accuracy of camera calibration. Then a new 3D reconstruction algorithm is proposed by adding the image selection step to 3D reconstruction. The image selection method uses structure-from-motion algorithm to estimate the position and attitude of each camera, first. Then the contributed value to 3D reconstruction of each image is calculated. Finally, images are selected according to the contributed value of each image and their effects on the contributed values of other images. Experimental results show that our image selection algorithm can improve the accuracy of camera calibration and the 3D reconstruction algorithm proposed in this paper can get better dense 3D models than the normal algorithm without image selection.     

Research on the magnetic resonance imaging brain tumor segmentation algorithm based on DO-UNet

With the social and economic development and the improvement of people's living standards, smart medical care is booming, and medical image processing is becoming more and more popular in research, of which brain tumor segmentation is an important branch of medical image processing. However, the manual segmentation method of brain tumors requires a lot of time and effort from the doctor and has a great impact on the treatment of patients. In order to solve this problem, we propose a DO-UNet model for magnetic resonance imaging brain tumor image segmentation based on attention mechanism and multi-scale feature fusion to realize fully automatic segmentation of brain tumors. Firstly, we replace the convolution blocks in the original U-Net model with the residual modules to prevent the gradient disappearing. Secondly, the multi-scale feature fusion is added to the skip connection of U-Net to fuse the low-level features and high-level features more effectively. In addition, in the decoding stage, we add an attention mechanism to increase the weight of effective information and avoid information redundancy. Finally, we replace the traditional convolution in the model with DO-Conv to speed up the network training and improve the segmentation accuracy. In order to evaluate the model, we used the BraTS2018, BraTS2019, and BraTS2020 datasets to train the improved model and validate it online, respectively. Experimental results show that the DO-UNet model can effectively improve the accuracy of brain tumor segmentation and has good segmentation performance.     

基于集成学习的强鲁棒性三维点云数据分类研究

目的通过三维扫描仪得到的点云数据往往存在很多异常值,例如噪点、遗失点和外部点等。在这些异常值存在的情况下,为了提高三维点云数据的分类精度,提出一种基于集成学习的强鲁棒性三维点云数据分类方法。方法提出一种基于最大投票法的集成学习思想,将2个深度神经网络的分类结果进行集成,从而提高网络的泛化性和准确性;采用全局特征增强和中心损失函数来优化神经网络结构,提高分类精度并增强鲁棒性。结果文中方法缩短模型训练时间至30个迭代次数,且在有噪点、丢失点和外部点的情况下分类精度均得到有效提升。结论提出的EL-3D算法在含有噪点、丢失点和外部点的情况下,鲁棒性效果要优于目前的点云分类方法。     

基于图像处理的外螺纹三维模型重构

提出了一种基于多角度序列图像特征实现外螺纹的三维模型重建的方法。首先在旋转平台上采集多角度序列螺纹件图像,然后对每帧图像进行特征点提取,将序列图像的特征点进行三维变换和插值,最终生成三维模型。实验结果表明,此算法能精确高精度地实现外螺纹三维模型重构。     

基于OCT成像的淡水无核珍珠内部缺陷自动检测方法

光学相干层析技术(OCT)作为一种高分辨率的无损光学检测手段,已被用于珍珠的内部质量检测。针对淡水无核珍珠质层内部缺陷检测的需求,提出一种通过光学相干层析图像实现淡水无核珍珠内部缺陷自动检测的方法。根据珠层灰度变化的特点,识别图像中缺陷区域的梯度特征和缺陷位置变化特征,并利用缺陷特征建立反向传播神经网络模型。实验中采集了内部无缺陷和内部有多种类型缺陷淡水无核珍珠的光学相干层析图像各20幅,对图像进行预处理并提取特征,利用K-means算法检测样本类型与所提取特征的匹配度,用特征与类型相匹配的样本特征训练反向传播神经网络模型,使用反向传播网络模型对淡水无核珍珠内部缺陷层进行分类识别。实验结果表明该方法提取特征的匹配度为92.5%,分类准确率达到100%,验证了该方法的可行性和有效性,提出的方法能够作为淡水无核珍珠内部缺陷识别和自动分类的有效手段。     

基于SIFT变换和RBF神经网络的彩色全景图拼接算法

目的 为了复原存在平移、色彩差异、旋转、形变等问题的全景图,提出一种结合SIFT（尺度不变特征变换）和RBF神经网络的彩色全景图拼接算法。方法 通过SIFT算法匹配出两子图中对应的特征点,利用仿射变换解决图像间的旋转和形变问题,采用RBF神经网络纠正子图的色彩差异,最后利用权值矩阵融合技术实现重叠区域的融合。结果 文中算法在拼接效果上优于其他算法,其拼接效果DoEM值为0.902,图像重叠区域过度平滑,有效地避免了融合区域的亮度块或亮度线。结论 该算法效果好,可解决全景图复原过程中多方面的难题。     

An Image Edge Detection Algorithm Based on Multi-Feature Fusion

Edge detection is one of the core steps of image processing and computer vision. Accurate and fine image edge will make further target detection and semantic segmentation more effective. Holistically-Nested edge detection (HED) edge detection network has been proved to be a deep-learning network with better performance for edge detection. However, it is found that when the HED network is used in overlapping complex multi-edge scenarios for automatic object identification. There will be detected edge incomplete, not smooth and other problems. To solve these problems, an image edge detection algorithm based on improved HED and feature fusion is proposed. On the one hand, features are extracted using the improved HED network: the HED convolution layer is improved. The residual variable convolution block is used to replace the normal convolution enhancement model to extract features from edges of different sizes and shapes. Meanwhile, the empty convolution is used to replace the original pooling layer to expand the receptive field and retain more global information to obtain comprehensive feature information. On the other hand, edges are extracted using Otsu algorithm: Otsu-Canny algorithm is used to adaptively adjust the threshold value in the global scene to achieve the edge detection under the optimal threshold value. Finally, the edge extracted by improved HED network and Otsu-Canny algorithm is fused to obtain the final edge. Experimental results show that on the Berkeley University Data Set (BSDS500) the optimal data set size (ODS) F-measure of the proposed algorithm is 0.793; the average precision (AP) of the algorithm is 0.849; detection speed can reach more than 25 frames per second (FPS), which confirms the effectiveness of the proposed method.     

基于自适应特征融合的多尺度相关滤波跟踪

为了减小目标跟踪中目标变形、光照影响、运动模糊以及目标旋转对跟踪效果的影响,在相关滤波KCF基础上,提出了一种基于自适应特征融合的多尺度相关滤波跟踪算法。首先,提取VGG19网络中conv2-2、conv3-4、conv5-4层的特征以及CN特征,并在conv2-2层加入CN特征;然后,将这3个特征分别代替HOG特征进行滤波学习,得到3幅响应图;进而对3幅响应图进行加权融合预测目标位置。最后,在尺度方面引入多尺度相关滤波器进行尺度的确定。该算法比KCF跟踪算法精确度和成功率分别提高了13.6%和11.8%。与现有的其他优异跟踪算法相比,该算法在应对运动模糊、背景杂乱、目标变形、平面旋转方面更具有较好的跟踪效果。     

基于卷积神经网络的超分辨率失真控制图像重构研究

目的 解决超分辨率图像重构模型中存在的功能单元之间关联性差,图像色度特征提取完整性不强、超分辨率重构失真控制和采样过程残差控制偏弱等问题。方法 通过在卷积神经网络模型引入双激活函数,提高模型中各功能单元之间的兼容连接性;引用密集连接卷积神经网络构建超分辨率失真控制单元,分别实现对4个色度分量进行卷积补偿运算;将残差插值函数应用于上采样单元中,使用深度反投影网络规则实现超分辨率色度特征插值运算。结果 设计的模型集联了内部多个卷积核,实现了超分辨率色度失真补偿,使用了统一的处理权值,确保了整个模型内部组成单元的有机融合。结论 相关实验结果验证了本文图像重构模型具有良好可靠性、稳定性和高效性。     

Electrical Capacitance Volume Tomography

A dynamic volume imaging based on the principle of electrical capacitance tomography (ECT), namely, electrical capacitance volume tomography (ECVT), has been developed in this study. The technique generates, from the measured capacitance, a whole volumetric image of the region enclosed by the geometrically three-dimensional capacitance sensor. This development enables a real-time, 3-D imaging of a moving object or a real-time volume imaging (4-D) to be realized. Moreover, it allows total interrogation of the whole volume within the domain (vessel or conduit) of an arbitrary shape or geometry. The development of the ECVT imaging technique primarily encloses the 3-D capacitance sensor design and the volume image reconstruction technique. The electrical field variation in three-dimensional space forms a basis for volume imaging through different shapes and configurations of ECT sensor electrodes. The image reconstruction scheme is established by implementing the neural-network multicriterion optimization image reconstruction (NN-MOIRT), developed earlier by the authors for the 2-D ECT. The image reconstruction technique is modified by introducing into the algorithm a 3-D sensitivity matrix to replace the 2-D sensitivity matrix in conventional 2-D ECT, and providing additional network constraints including 3-to-2-D image matching function. The additional constraints further enhance the accuracy of the image reconstruction algorithm. The technique has been successfully verified over actual objects in the experimental conditions     

基于改进YOLOv5的试剂卡印刷缺陷检测算法

目的 针对试剂卡生产企业采用人工分选印刷缺陷的试剂卡存在效率低、成本高、易漏检的问题，提出一种基于深度神经网络YOLOv5s的改进试剂卡印刷缺陷检测算法YOLOv5s-EF。方法 通过图像预处理算法获得高质量的缺陷图像数据集，在YOLOv5s的主干特征提取网络中添加高效通道注意力（Efficient Channel Attention, ECA）机制，增强特征图中重要特征的表示能力；引入焦点损失函数（Focal Loss）来缓解正负样本不均衡的影响；结合印刷区域的定位结果，二次精确定位并构建方位特征向量，提出一种特征向量相似度匹配方法。结果 实验结果表明，本文提出的试剂卡印刷缺陷检测算法在测试集上的检测平均准确度可以达到97.3%，速度为22.6帧/s。结论 相较于其他网络模型，本文提出的方法可以实现对多种印刷缺陷的识别与定位，模型具有较好的检测速度和鲁棒性，有利于提高企业生产的智能化水平。     

Comparative analysis of Alzheimer's disease classification by CDR level using CNN,feature selection,and machine-learning techniques

Magnetic resonance imaging (MRI) of brain needs an impeccable analysis to investigate all its structure and pattern. This analysis may be a sharp visual analysis by an experienced medical professional or by a computer aided diagnosis system that can help to predict, what may be the recent condition. Similarly, on the basis of various information and technique, a system can be designed to detect whether a patient is prone to Alzheimer's disease or not. And this task of detection of abnormalities at an initial stage from brain MRI is a major challenge in the field of neurosciences. The main idea behind our research is to utilize the deep layers feature extraction benefited from deep neural network architecture, without extensive hardware resource training, and classifying the image on a basis of simple machine-learning algorithm with selected best features in order to reduce work load, classification error and hardware utilization time. We have utilized convolution neural network (CNN) layer using similar architecture like that of Alexnet with some parametric change, for the automatic extraction of features of images obtained from slice extraction of whole brain MRI whereas 13 manual features based on gray level co-occurrence matrix were also extracted to test the impact of this features on ranking. If we had only classified using CNN network, the misclassification rate was much higher. So, feature selection is achieved with feature ranking algorithms like Mutinffs, ReliefF, Laplacian and UDFS and so on and also tested with different machine-learning techniques like Support Vector Machine, K-Nearest Neighbor and Subspace Ensemble under different testing condition. The performance of the result is satisfactory with classification accuracy around 98% to 99% with 7:3 ratio of random holdout partition of training to testing image sets and also with fivefolds of cross-validation on the same set using a standardized template.     

基于多级特征融合的体素三维目标检测网络

目的 为精确分析点云场景中待测目标的位置和类别信息,提出一种基于多级特征融合的体素三维目标检测网络。方法 以2阶段检测算法Voxel?RCNN作为基线模型,在检测一阶段,增加稀疏特征残差密集融合模块,由浅入深地对逐级特征进行传播和复用,实现三维特征充分的交互融合。在二维主干模块中增加残差轻量化高效通道注意力机制,显式增强通道特征。提出多级特征及多尺度核自适应融合模块,自适应地提取各级特征的关系权重,以加权方式实现特征的强融合。在检测二阶段,设计三重特征融合策略,基于曼哈顿距离搜索算法聚合邻域特征,并嵌入深度融合模块和CTFFM融合模块提升格点特征质量。结果 实验于自动驾驶数据集KITTI中进行模拟测试,相较于基线网络,在3种难度等级下,一阶段检测模型的行人3D平均精度提升了3.97%,二阶段检测模型的骑行者3D平均精度提升了3.37%。结论 结果证明文中方法能够显著提升目标检测性能,且各模块具有较好的移植性,可灵活嵌入到体素类三维检测模型中,带来相应的效果提升。     

基于图像监测的微量移液技术的研究

为了提高生化分析仪微量移液的可靠性,提出了一种基于图像分割法的移液故障判断和移液量检测系统。基于STM32微控制器,利用速度位置双闭环PID （proportion-integral-derivative,比例-积分-微分）算法控制微量移液系统的步进电机,设计了生化分析仪的自动移液控制系统。采集移液过程图像作为样本图像,建立了移液数据集。分别对剪枝前后的U-Net神经网络模型进行训练,并比较模型的计算量、参数量和平均交并比。通过剪枝后U-Net模型对移液区域的分割和处理,实现移液系统的故障判断,并结合移液吸头的几何特征建立移液体积模型,以计算移液体积。进行了基于图像分割法的微量移液实验,对移液误差进行分析,并利用最小二乘法、亚像素角点检测和极限学习机进行误差补偿。结果表明,剪枝后U-Net神经网络模型的计算量、参数量和平均交并比分别下降了47.30%,93.99%和0.61%,模型运行效率显著提升。针对10,50,100 μL的检定点,微量移液系统的移液精度分别达到1.72%、1.36%和1.39%,满足微量移液系统的精度设计要求。基于图像分割法的移液监测系统能够有效判断移液故障并检测移液体积。研究结果对微量移液技术的发展起到一定的推动作用。