基于改进神经网络算法的蔬菜图像识别

文章提出了改进神经网络算法,建立了径向基函数神经网络模型,包括梯度下降方法求解权重参数,增大邻域半径的均值聚类方法求取隐函数中心值,利用相邻聚类中心获得核宽度,通过量子遗传算法删除冗余权重和神经元;提取了蔬菜图像的特征,并给出了算法流程。仿真试验表明,试验算法对蔬菜图像的形状特征平均识别率为97.56%,纹理特征平均识别率为95.60%,颜色特征平均识别率为93.25%,训练时间平均为5.83s、识别时间平均为2.18s,优于其他算法。     

基于改进AlexNet模型的抓毛织物质量检测方法

针对传统图像识别方法对抓毛织物表面特征难以提取且识别准确率低的问题,提出了一种改进AlexNet模型的抓毛织物质量检测方法,通过数据增强方法对抓毛织物数据进行扩充,构建卷积神经网络对抓毛织物的样本特征进行提取,利用SGDM、RMSProp、Adam优化算法和改变学习率相结合的实验方法,采用全新学习与迁移学习两种算法对抓毛织物图像数据集进行训练,在训练完成后,分别利用卷积神经网络的不同深度池化层提取抓毛织物样本的特征作为输入,将提取到的抓毛织物特征拟合支持向量机(SVM)分类器,最后对输入的抓毛织物图像进行分类。实验结果表明:使用卷积神经网络方法能够增加卷积层对抓毛织物表面特征的提取能力,获得具有较高分辨力的图像特征,通过数据增强和SGDM算法训练的模型,提取网络pool5层特征拟合SVM分类器,识别准确率明显提高。基于改进AlexNet模型的抓毛织物质量检测方法能够提取抓毛织物表面特征且识别率高。     

基于PSO-BP神经网络的数码印花颜色预测模型研究

面向定制时代纺织工业数码印花的关键技术,根据数码印花的颜色复现特征,提出一种基于粒子群优化算法（PSO）优化三层BP神经网络的颜色预测模型,通过权值和阈值优化解决了BP神经网络易陷入局部极小值的难题,有效提高了数码印花颜色预测精度。     

基于稀疏表示分类的家具表面缺陷检测

针对家具表面的死节缺陷,提出一种基于图像颜色特征和稀疏表示分类的家具表面缺陷检测方法。首先将RGB转换为HSI图像,分别提取HSI、RGB空间颜色分量,将各颜色分量灰度图分成一系列小块,提取每块子图的特征值,选取部分缺陷与背景子图的特征值作为训练样本集,利用稀疏表示分类算法得到家具表面的缺陷目标,最后经过腐蚀膨胀消除缺陷边界小点使缺陷边界平滑,实现家具表面的缺陷检测。实验结果显示,该方法能够有效的检测出家具表面死节缺陷,为现代家具表面缺陷检测提供准确、快速的识别算法。     

基于颜色特征和BP神经网络的大米加工精度判别方法研究

提出了基于颜色特征和BP神经网络判别大米加工精度的方法。设计了基于机器视觉的大米加工精度检测装置采集大米图像,利用图像处理技术对获取的大米图像进行预处理,提取大米籽粒目标图像;在大米籽粒腹部确定半径为R的圆形区域作为颜色特征值提取区域,将颜色特征值提取区域按面积平均分成5个同心圆子区域,提取每个子区域的R、G、B颜色值,并将颜色值转成色调H值作为描述大米籽粒表面加工精度的颜色特征值,以5个颜色特征值作为输入值,采用BP神经网络对大米的加工精度进行检测。试验结果表明:该方法对4种不同加工精度大米样品籽粒检测的平均准确率为92.17%。     

计算机视觉用于猪肉新鲜度检测的颜色特征优化选取

在以计算机视觉为基础,并利用神经网络预测猪肉通脊新鲜度时,选择合适的颜色特征参数和神经网络模型是提高其预测准确性的关键之一。文中提出了一种猪肉新鲜度等级预测时颜色特征参数和神经网络优化选取的方法,利用图像处理的方法提取通脊表面的颜色特征参数,组合成RGB-HIS、RGB-L*a*b*、rgb-HIS、rgb-L*a*b*及HIS-L*a*b*五类特征参数组合,并利用BP(back propagation,BP)和SVM(support vector machine,SVM)神经网络构造各类新鲜度等级预测模型。结果表明:SVM和BP的平均预测准确率分别为91.11%和84.44%,且rgb-HIS特征参数组合的BP与SVM预测模型的预测准确率最高,分别为88.89%和95.56%。因此,提取通脊表面r、g、b、H、I、S均值作为颜色特征向量,且选择SVM神经网络来构造新鲜度预测模型可显著提高预测结果的准确性。     

基于模拟退火算法改进遗传算法的织物智能配色

为提高传统配色方法及现有配色算法的配色精度、效率及泛化能力,构建了基于BP神经网络的遗传算法和模拟退火算法相结合的织物智能配色模型,利用BP神经网络预测颜色,将训练好的BP神经网络与CIEDE2000色差公式结合作为遗传算法的适应度函数,用模拟退火算法改进的基于BP神经网络的遗传算法预测颜色配方,并根据预测的配方对涤纶织物进行染色实验,计算实验色差。结果表明:模拟退火算法优化的基于BP神经网络的遗传算法配色模型只需经过80次迭代即可收敛,预测颜色的理论色差均值为0.165,染色实验色差均值为0.289,配方绝对误差平均值为0.010 7;验证样本的理论色差均值为0.240,染色实验色差均值为0.437。该算法可实现织物的智能配色。     

基于空洞卷积神经网络的艺术图像风格重建算法

针对普通卷积神经网络提取图像特征能力表现欠佳的问题,提出一种利用空洞卷积神经网络对艺术图像和照片做融合重建的算法。通过设计不同融合程度的损失函数实现对艺术图像纹理信息和照片中内容表现的特征提取,利用随机梯度下降算法对整体的损失函数做迭代改进,实现艺术图像风格和内容的融合重建。实验结果表明,文中方法相比于使用普通卷积神经网络融合特征的方法具有更高的可靠性和更优秀的表现。     

利用颜色和纹理进行图像检索的算法分析

自从上世纪70年代,人们就开始了对图像检索技术的研究,但是只是基于文本方面的,而且仅仅是利用文本的方式对图像的特征进行简单的描述。后来,人们不在满足于简简单单的图像检索方法,进而又开始进行了更深层次的开发,这样就出现了利用颜色和纹理对图像的检索和分析的方法,这种检索的方法是基于图像的内容进行检索的一种方法。这种检索方法只是对文本进行检索的方法更为精确。所以人们开始使用这种利用颜色和纹理进行检索的方法,进而研究了这种检索方法的检索算法。在利用颜色和纹理进行图像检索的算法中,要对图像进行分割,利用分割好的图像进行颜色平均值的换算,再把颜色平均值进行低频分量,把它作为颜色的特征。在根据边缘直方图来提取检索内容的纹理特征,最后根据这两个特征进行图像的检索。在本文中关于检索的算法的实践性进行了说明,这样的检索算法可以有效的提高图像的准确率。     

改进频率调谐显著算法在疵点辨识中的应用

为提高疵点检测效率和准确率,提出用改进频率调谐显著（FT）算法替代 Gabor 小波方法预处理疵点图像,强化疵点特征向量灵敏度。分析了FT 算法中高斯滤波器模板、Lab 颜色空间、高斯滤波图像中椒盐噪声和 HSV 颜色空间不同通道取值范围不一致对疵点识别的影响,并提出了相应改进方法。利用改进 FT 算法进行图像显著处理;使用灰度共生矩阵方法对疵点显著图进行特征提取;利用概率神经网络分类器分类,检测是否存在疵点。对 2 种不同纹理面料的检测结果表明：改进 FT 算法较改进前计算时间增加约8%,但疵点检测准确率提高18% ～25%;与 Gabor 小波相比,检测准确率基本持平,但计算时间缩短约70%。     

Discriminating and elimination of damaged soybean seeds based on image characteristics

To identify and eliminate damaged soybean seeds, images of Kaiyu 857 soybean seeds including those with insect damage, mildew, and other defects were acquired with an intelligent camera. After splitting the kernels from the background through using the data fusion, morphological corrosion expansion and a series of image processing algorithms, we extracted eight shape features, three color features and three texture features as the input layer to set up a BP neural network classification model with an average recognition accuracy of 97.25%. The identifying and eliminating device was tested five times with a mixture of 1000 differently damaged soybeans of seeds. The average accuracy rates of identification and elimination for normal, mildewed, insect-damaged, skin-damaged, broken and partly defective kernels reached 99.24%, 98.2%, 96.4%, 85.6%, 92.4% and 85.2% respectively. The efficient processing speed of the device reached 125 grains per minute. The results are of significance for the development of precise selection systems for soybeans or other crop seeds.     

Evaluation of pork color by using computer vision

The objective of this study was to determine the potential of computer vision technology for evaluating fresh pork loin color. Software was developed to segment pork loin images into background, muscle and fat. Color image features were then extracted from segmented images. Features used in this study included mean and standard deviation of red, green, and blue bands of the segmented muscle area. Sensory scores were obtained for the color characteristics of the lean meat from a trained panel using a 5-point color scale. The scores were based on visual perception and ranged from 1 to 5. Both statistical and neural network models were employed to predict the color scores by using the image features as inputs. The statistical model used partial least squares technique to derive latent variables. The latent variables were subsequently used in a multiple linear regression. The neural network used a back-propagation learning algorithm. Correlation coefficients between predicted and original sensory scores were 0.75 and 0.52 for neural network and statistical models, respectively. Prediction error was the difference between average sensory score and the predicted color score. An error of 0.6 or lower was considered negligible from a practical viewpoint. For 93.2% of the 44 pork loin samples, prediction error was lower than 0.6 in neural network modeling. In addition, 84.1% of the samples gave an error lower than 0.6 in the statistical predictions. Results of this study showed that an image processing system in conjunction with a neural network is an effective tool for evaluating fresh pork color.     

基于图像处理的烟叶烘烤阶段判别模型优选

[背景和目的]烟叶烘烤阶段的自动判别是建立智能化烟叶烘烤系统的重要环节.为实现烘烤阶段的精确识别和操控,提升烟叶烘烤的精准度.[方法]提取烘烤过程中整夹烟叶图像的11种颜色特征和8种纹理特征,分别对颜色特征和纹理特征进行变量聚类分析,以10为距离,将提取的颜色特征和纹理特征各分为2类.利用相关性分析筛选出每类特征中与烘...     

基于BP神经网络的色纺纱配色

针对传统配色模型实用性差的问题,利用神经网络强大的非线性映射能力,探讨基于人工神经网络的色纺纱配色方法。构建了色纺纱BP神经网络配色模型,研究了多种BP算法,隐含层节点数对仿真效果及泛化能力的影响。实验结果表明:基于BP 神经网络的色纺纱配色方法可以实现色纺纱反射率与配方之间的非线性映射,新型算法（Levenberg-Marquardt、拟牛顿、共轭梯度算法）在迭代次数和仿真时间上有较大的优势,隐含层节点数对仿真结果影响较小,训练样本的平均预测色差DEcmc(2:1)为0.18,但训练样本范围外的预测色差较大,因此提高该神经网络的泛化性能是下一步研究的关键。     

基于高光谱图像光谱变量和颜色特征的霉变玉米籽粒识别

目的:准确识别霉变玉米籽粒。方法:基于高光谱图像光谱变量和颜色特征建立霉变玉米籽粒识别的新方法。先对玉米籽粒图像进行图像分割和光谱变量、颜色特征提取,并根据颜色特征生成颜色直方图;将光谱变量和颜色直方图特征组成特征集合;通过距离函数对特征集合中所有特征的分析确定霉变玉米籽粒所属类别。结果:所提方法对霉变玉米籽粒类别的最大平均识别偏差为1.12,最佳平均识别准确率为97.59%;与基于高光谱图像+随机蛙跳+极限学习机的方法、基于高光谱图像+稀疏自动编码器+卷积神经网络的方法、基于高光谱图像+蚁群优化+BP神经网络的方法相比,研究所提方法对霉变玉米籽粒类别的识别准确率明显提高。结论:该方法可实现被测玉米籽粒样品是否霉变以及霉变程度的准确判断。     

基于神经网络的男西装图像情感语义识别

为了实现基于内容的男西装图像情感语义识别,需要把男西装图像的低层特征映射到情感语义空间。在构建出的2维图像情感因子空间和男西装图像视觉特征（10维亮度一冷暖模糊直方图;7维的饱和度一冷暖模糊直方图+色彩对比度值的综合特征）的基础上,本文通过机器学习（BP神经网络）实现了男西装图像的低层特征到情感语义因子空间的映射,根据图像低层颜色特征可以自动完成图像情感因子值和情感描述值的计算,并把识别后的新图像数据自动加入到图像数据库中。实验证明,BP神经网络方法能较好的实现基于内容的男西装图像情感语义的识别。     

Determination of animal skeletal maturity by image processing

Color image features were computed to characterize the skeletal maturity of beef carcasses based on cartilage ossification in the thoracic vertebrae. A trained neural network was tested for predicting USDA beef maturity grades from image features of ossification. A feature curve was defined to characterize the color variations of an isolated cartilage-bone object. Both RGB and HSL color systems were used to derive image features. The maturity grades were assigned by an official USDA grader. Two sets of samples were obtained from two different meat-processing plants. The first set contained samples of only A and B maturity grades whereas the second set had all five maturity classifications (A through E). The hue value was the most useful color feature. The mean hue values of cartilage differed (P<0.05) among the maturity grades and the feature curve based on the hue value was used as neural network input for maturity prediction. The accuracy of prediction was 75% for the first set of samples and 65.9% for the second set of samples. The results data show the potential of computer vision techniques for beef maturity assessment.     

Identifying Potato Varieties Using Machine Vision and Artificial Neural Networks

The objective of this study is to develop a method for identifying and discriminating 10 potato varieties by combining machine vision and artificial neural network methods. The potato varieties include Agria, Savalan, Florida, Fontaneh, Natasha, Verona, Karso, Elody, Satina, and Emrad. A total number of 72 characteristic parameters specifying color, textural, and morphological features are found among these varieties. By using principal component analysis, 16 principal features are selected for identifying and discriminating potato varieties. The data obtained from image processing were classified using linear discriminant analysis and non-linear artificial neural network method. The accuracy of discriminant analysis were 73.3, 93.3, 73.3, 40, 73.3, 73.3, 66.7, 80, 40, and 53.3%, respectively, for the varieties used in this study. The classification accuracy was improved by 100% for all the varieties using neural network analysis and the correct classification ratio was 100% using this method. It is revealed from the results that machine vision technique and neural network analysis could identify potato varieties with acceptable accuracy.     

BP 神经网络在大豆油酸价近红外光谱检测中的应用

对于大豆四级油,采用BP 神经网络对其近红外光谱数据建模,对系统的结构及参数选取进行了分析,对样本训练集的设计和网络输入端的主因子方面进行了处理。对于其他的多变量建模方法,分析了其对近红外光谱有用信息的提取作用。结果显示：多元线性回归、主成分回归和偏最小二乘法对大豆四级油酸价预测的标准偏差分别为0.1472%、0.1801% 和0.1576%,BP 神经网络的预测标准偏差为0.1387%。