卷积神经网络金相组织自动识别

为了降低人工分辨金相组织图像类别的误差率,提高分辨效率,采用卷积神经网络模型对金相组织图像进行自动辨识。对制备金相样块所得铁素体与马氏体两种金相组织图像进行分析,提出符合金相组织图像分布特征的预处理方案。通过采用图像尺寸归一化、灰度值归一化以及高斯平滑处理等方法,对原始金相组织图像进行预处理,建立金相组织图像数据集。针对建立的铁素体和马氏体金相组织图像数据集,提出了适合金相组织图像辨识的改进模型,分别记为LeNet-MetStr模型、AlexNet-MetStr模型和VGGNet-MetStr模型。对3种改进卷积神经网络进行模型训练及分析,结果表明VGGNet-MetStr模型对2种金相组织图像自动辨识具有更高的准确度。     

Modulation classification in the presence of adjacent channel interference using convolutional neural networks

This paper investigates a vital issue in wireless communication systems, which is the modulation classification. A proposed framework for modulation classification based on deep learning (DL) is presented in the presence of adjacent channel interference (ACI). This framework begins with the generation of constellation diagrams from the received data. These constellation diagrams are fed to convolutional neural networks (CNNs) for modulation classification. The objective of this process is to eliminate the manual feature extraction from the received data and make feature extraction process as a built‐in step with CNNs. Three types of CNNs are considered in this paper and compared for this objective. These types are AlexNet, VGG‐16, and VGG‐19. The proposed classifier is applied on Rayliegh and Rician fading channels.     

多感知兴趣区域特征融合的图像识别方法

针对自然图像识别过程中不同深度学习模型关注兴趣区域不同的现象,本文引入深度卷积神经网络融合机制,结合深度迁移学习方法,给出了一种基于多感知兴趣区域特征融合的图像识别方法。本文将迁移学习方法引入牛津大学视觉组网络模型(visual geometry group network,VGGNet)和残差网络模型(residual network,ResNet),通过对单个分类模型进行热力图可视化及特征可视化,得到了不同网络模型关联的特征区域不一样的结论。然后在此基础上分别设计特征拼接、特征融合加特征拼接及融合投票方法将不同模型特征进行融合,得到3种新的融合模型。实验结果表明,本文方法在Kaggle数据集上的识别准确率高于VGG-16、VGG-19、ResNet-50、DenseNet-201模型。     

基于VGGNet卷积神经网络的加密芯片模板攻击新方法

针对传统模板分析在实际攻击中的难解问题，重点研究了在图像识别领域具有优异特征提取能力的VGGNet网络模型，提出了一种基于VGGNet网络模型的模板攻击新方法。为了防止信号质量对模型准确率带来较大影响，采用相关性能量分析方法对采集到的旁路信号质量进行了检验；为了适应旁路信号数据维度特征，对网络模型结构进行适度调整；在网络训练的过程中，对梯度下降速率较慢、梯度消失、过拟合等问题进行了重点解决，并采用五折交叉验证的方法对训练好的模型进行验证。最终实验结果表明，基于VGGNet模型的测试成功率为92.3%，较传统的模板攻击效果提升了7.7%。     

融合头部姿态和面部表情的互动式活体检测

为了阻挡人脸识别系统中的照片及视频攻击,提出了一种将头部姿态和面部表情融合的互动式活体检测算法。首先,对VGGNet的卷积核数目、网络层数、正则化等进行了调整优化,构建了一个多层卷积的头部姿态估计网络;其次,引入全局平均池化、局部响应归一化和卷积替代池化等方法对VGGNet进行改进,构建了一个表情识别网络;最后,融合上述两个网络实现了互动式活体检测系统,对用户发出随机指令实时完成活体检测。实验结果表明,所提出的头部姿态估计网络和表情识别网络分别在CAS-PEAL-R1数据集和CK+数据集上取得了99.87%和99.60%的准确率,而活体检测系统的综合准确率达到了96.70%,运行速度达到了每秒20~28帧,在实际应用中泛化能力突出。     

基于VGGNet和多谱带循环网络的高光谱人脸识别系统

为了提高光谱人脸数据表征人脸特征的有效性，提出一种基于VGGNet和多谱带循环训练的高光谱人脸识别方法。首先，在光谱人脸图像的预处理阶段，采用多任务卷积神经网络（MTCNN）进行高光谱人脸图像的精确定位，并利用混合通道的方式对高光谱人脸数据进行增强；然后，基于卷积神经网络（CNN）结构建立一个面向高光谱人脸识别的VGG12深度网络；最后，基于高光谱人脸数据的特点，引入多谱带循环训练方法训练建立的VGG12网络，完成最后的训练和识别。在公开的UWA-HSFD和PolyU-HSFD高光谱人脸数据集的实验结果表明，所提方法取得了比其他深度网络（如DeepID、DeepFace、VGGNet）更好的识别性能。     

基于VGGNet颜色识别的车辆检索系统设计

The Convolution Neural Network(CNN) in Deep Learning has a strong anti-jamming ability for image translation, rota- tion and other transformations. Compared with the traditional vehicle recognition technology, it can extract deeper and richer image information. Based on the VGGNet structure and simulating the order of human eyes' perception of vehicle characteristics, this pa- per designs a hierarchical retrieval system for vehicle image database. Firstly, a CNN which can recognize eight kinds of colors is constructed and trained to recognize the color of the target vehicle. Then, SIFT and LBP features are combined to match and retrieve the same color candidate vehicle database. The hierarchical retrieval mode of the system can effectively reduce the scope of retrieval and improve the efficiency of retrieval. The fusion of multi features can also guarantee the extraction of enough image information and ensure the accuracy of retrieval.     

多尺度双通道卷积神经网络下的刺绣模拟

目的 针对现有刺绣模拟算法中针线感不强、针线轨迹方向单一等问题,提出了一种基于多尺度双通道卷积神经网络的刺绣模拟算法。方法 1）搭建多尺度双通道网络,选取一幅刺绣艺术作品作为风格图像,将MSCOCO（microsoft common objects in context）数据集作为训练集,输入网络得到VGG（visual geometry group）网络损失和拉普拉斯损失;2）将总损失值传回到网络,通过梯度下降法更新网络参数,并且重复更新参数直到指定的训练次数完成网络训练;3）选取一幅目标图像作为刺绣模拟的内容图像,输入训练完成的网络,获得具有刺绣艺术风格的结果图像;4）使用掩模图像将得到的结果图像与绣布图像进行图像融合,即完成目标图像的刺绣模拟。结果 本文算法能产生明显的针线感和多方向的针线轨迹,增强了刺绣模拟绘制艺术作品的表现力。结论 本文将输入图像经过多尺度双通道卷积神经网络进行前向传播,并使用VGG19、VGG16和拉普拉斯模块作为损失网络进行刺绣模拟。实验结果表明,与现有卷积神经网络风格模拟算法对比,本文提出的网络能够学习到刺绣艺术风格图像的针线特征,得到的图像贴近真实刺绣艺术作品。     

基于VGGNet卷积神经网络的蓝印花布纹样分类

为了数字化传承与创新传统的蓝印花布纹样,需 要将蓝印花布纹样进行分类。为此,提出一种改进的VGGNet卷 积神经网络模型的纹样分类方法。首先,采集原始的蓝印花布图案,通过图像增强技术扩充 样本,形成训练数据集。其次, 改进经典的VGGNet 16卷积神经网络结构,增加卷积组及调整网络参 数,增加丢弃层。同时,分析、验证训练优化策略对 蓝印花布纹样分类的影响。最后,利用训练集及验证集中的图像样本,通过自动学习获取网 络模型参数,得到纹样分类的最 佳网络模型并获得较为理想的分类结果。实验结果显示,改进的卷积神经网络模型针对5类 蓝印花布纹样进行分类训练,其 平均分类准确率达89.73%,为蓝印花布纹样的继承和创新研究提供了 新思路。     

Image-Based Automatic Diagnostic System for Tomato Plants Using Deep Learning

Tomato production is affected by various threats, including pests, pathogens, and nutritional deficiencies during its growth process. If control is not timely, these threats affect the plant-growth, fruit-yield, or even loss of the entire crop, which is a key danger to farmers’ livelihood and food security. Traditional plant disease diagnosis methods heavily rely on plant pathologists that incur high processing time and huge cost. Rapid and cost-effective methods are essential for timely detection and early intervention of basic food threats to ensure food security and reduce substantial economic loss. Recent developments in Artificial Intelligence (AI) and computer vision allow researchers to develop image-based automatic diagnostic tools to quickly and accurately detect diseases. In this work, we proposed an AI-based approach to detect diseases in tomato plants. Our goal is to develop an end-to-end system to diagnose essential crop problems in real-time, ensuring high accuracy. This paper employs various deep learning models to recognize and predict different diseases caused by pathogens, pests, and nutritional deficiencies. Various Convolutional Neural Networks (CNNs) are trained on a large dataset of leaves and fruits images of tomato plants. We compared the performance of ShallowNet (a shallow network trained from scratch) and the state-of-the-art deep learning network (models are fine-tuned via transfer learning). In our experiments, DenseNet consistently achieved high performance with an accuracy score of 95.31% on the test dataset. The results verify that deep learning models with the least number of parameters, reasonable complexity, and appropriate depth achieve the best performance. All experiments are implemented in Python, utilizing the Keras deep learning library backend with TensorFlow.