基于图像识别的苹果叶片病害识别模型对比研究

为实现苹果叶片病害图像自动识别,展开苹果叶片病害识别模型研究。通过整理网络开源植物病害数据,获取苹果赤霉病、苹果雪松锈病和苹果灰斑病3种苹果病害叶片图像,以1种健康叶片图像作为研究对象,随机抽取4433张图像建立数据集用于模型训练,采用离线增强和在线增强2种手段对数据进行预处理,扩充图像样本并保证各类样本均衡。在Resnet 50、Mobilenet v2、Vgg16、Vgg19、Inception v3等5种预训练模型的基础上,对迁移模型进行一系列的参数调整。5种模型训练比对结果表明,优化后的Resnet50模型能够达到0.9770的准确率。优化后的训练模型具有识别速度快、准确率提高的特点,可以准确、快速地识别出病害类型,为植物病害的自动诊断提供支撑。     

Face Recognition Based on Weighted PCA

This paper proposes a face novel recognition method based on Weighted PCA, which combined weighted features with PCA. At first, we calculate the weighted subspace for each class by minimizing the weighted reconstruction error. Then a test example is classified by the distance from weighted subspace. The experiments on Cambridge ORL database show that our Weighted PCA method can improve the recognition rate significantly without increasing the computation, when compared with PCA.     

Study on ANN Based Character Recognition

The theory of Artificial Neural Networks and image processing are analyzed. ANN theory is also investigated . A Back Propagation network is used for classification. Based on the ANN and feature extraction, a character recognition and entry system is developed. Its drawbacks are discussed. Some solutions are presented. An odd feature database is constructed to promote the performance for learning. A 3 layer feedforward network has been proved that it has the ability to identify hand written characters.     

Feature Recognition Based on Vector Picture

The technologyof recognition and understanding architecture engineering drawings by computer is a study hotspot in the field that computer isapplied in engineering research. Based on the characteristicof architectural ichnographyand criterionof architectural design fires prevention, the recognizing identification method and main program based on DXF format is proposed. Experimental results show that it has a preferableautomatic recognition results on the architecture ichnography,and effectivelyimproves the work efficiency.     

A Character Recognition Based on Feature Extraction

This thesis analyzes the theory of image processing and feature extraction, character feature is investigated in this thesis, including width, intersection, chain-code, the way of structure of character is used, a hand-written character recognition and entry system is developed. It is proved feature is simple and definite.     

玉米–大豆带状套作行距配置对作物生物量、根系形态及产量的影响

空间配置是影响间作套种作物生长和产量构成的关键因素之一。本研究固定玉米–大豆套作带宽200 cm,玉米采用宽窄行种植,设置4个玉米窄行行距为20 cm(A1)、40 cm(A2)、60 cm(A3)和80 cm(A4)套作处理,2个玉米和大豆净作对照处理,研究行距配置对套作系统中玉米和大豆生物量、根系及产量的影响。结果表明,套作大豆冠层光合有效辐射和红光/远红光比值均低于净作,且随着玉米窄行的增加而降低。套作系统中大豆地上地下生物量、总根长、根表面积和根体积从第三节龄期(V3)到盛花期(R2)逐渐增加,但随着玉米窄行的增加而降低。套作玉米地上地下生物量从抽雄期到成熟期逐渐增加,根体积却逐渐降低,但这些参数随玉米窄行的变宽而增加。玉米和大豆在带状套作系统中产量均低于净作,且随玉米窄行的变宽,玉米产量逐渐增加,2012和2013两年最大值平均为6181 kg hm–2,而大豆产量逐渐降低,两年最大值平均为1434 kg hm–2,产量变化与有效株数和粒数变化密切相关。此外,玉米–大豆带状套作群体土地当量比(LER)大于1.3,最大值出现在A2处理,分别为1.59(2012年)和1.61(2013年),且最大经济收益也出现在A2处理(2年每公顷平均收益为1.93万元)。因此,合理的行距配置对玉米–大豆带状套作系统中作物的生长、产量构成和群体效益具有重要的作用。     

Adaptive Simulation Algorithm Based on Fuzzy Stiffness Recognition

Adaptive simulation algorithm is an effective and practical method for simulating stiff system,and the key to realize this algorithm is stiffness recognisation.This paper,based onthe previous recognizing methods,introduces fuzzy set theory in them to develop a more accurate approach which is successfully used in adaptive simulation algorithm.     

基于机器视觉技术的黄瓜叶片白粉病识别

黄瓜叶片白粉病染病程度的判定,对于确定病灾预防措施意义重大。笔者采用机器视觉技术对叶片病斑进行有效识别,借助人工神经网络完成对叶片染病状态的模式分类。根据病斑的规模及面积分布提出了白斑区域面积比、平均白斑面积、白斑覆盖率等特征参数,借助这3个特征参数实现了叶片染病程度的定量分析,并借助人工神经网络完成了对叶片染病状态的模式分类。4类白斑叶片的正确识别率分别为88%、91%、92%、94%。     

基于改进的 AlexNet 卷积神经网络的植物叶片识别

针对多类植物叶片识别准确率和计算效率低问题,通过残差连接卷积层,构建了一种基于残差连接的Alexnet卷积神经网络。对卷积层的输入数据进行批归一化,进而加速网络的收敛速度。利用两种不同的全局池化算法既能减少特征图数目又能提高算法识别植物叶片的准确性和计算效率。对于不同种类植物叶片,实验验证残差连接的Alexnet卷积神经网络能够进行高效识别。     

Disign and Implementation of Speaker Recognition System Based on FPGA

Aiming at the shortcoming of low recognition and training speed in embedded speaker - recognition system based on DS Phard-core processor, a new scheme of system based on FPGA and vector quantization principle is presented. In the speaker-recognition system based on the vector quantization and generation algorithm, a fitness parallel process hardware structure is presented by the scheme which consumes much less time than software processing while getting the fitness. The test shows that the system uses this method obtains both high recognition rate and higher speed in training and recognition.     

Pd Pattern Recognition Based on Linear Discriminant Analysis in GIS

According to the character of PD in GIS, the authors design four kinds of GIS defection models. The GIS gray intensity images are constructed based on mass specimens gathered by the ultra - high frequency and high speeds systems, Aiming at the PD characteristics and its defections, A PCA-FDA method is put forward based on PD images. The principal component analysis is employed to condense the dimension of PD images, then the optimal sets of statistically uncorrelated discriminant vectors are extracted, and the minimum distance classifier is constructed as classifier. The identified results show that this method can effectively elevated the discrimination of the four kinds of defects in GIS PD.     

利用机载激光雷达的林木识别与参数反演

为了提高利用机载激光雷达林木识别精度,以全波形激光雷达分解数据为基础,首先利用控制标记分水岭算法对树冠高程模型进行分割,初步确定单株木位置;然后结合单株木结构特征,在三维空间中利用马尔可夫随机场进行单株木点云分割;最后,使用9个样地实测数据对激光雷达反演的林木参数进行回归分析验证。结果表明：单株木识别率为76%,位置误差均值和方差分别为0.67、0.19 m,单株木树高、树冠直径和胸高径的RMSE分别为1.03 m（4.57%）、0.56 m（10.48%）、3.01 cm（11.01%）,样地断面积和材积的RMSE分别为2.42 m2/hm2（8.11%）和17.83 m3/hm2（9.11%）。本研究能有效提高单株木点云分割精度,能满足单株木和林分参数反演要求,提高林业调查的自动化程度。     

基于粒子群算法和支持向量机的黄花菜叶部病害识别

使用数字图像处理技术,以黄花菜叶部病害图像为识别对象,基于Lab空间和K-means聚类算法分割病害区域,提取目标区域的颜色特征、方向梯度直方图(histogram of oriented gradient,HOG)特征和形状特征,分别建立单一特征模型和特征融合模型,采用粒子群(particle swarm optimization,PSO)算法通过交叉验证优化支持向量机(support vector machine,SVM)模型的惩罚因子和核参数,建立基于PSO-SVM的多特征融合分类模型识别黄花菜病害。基于SVM的多特征融合分类模型识别率高于单一特征分类模型,识别率可达为81.67%;基于PSO-SVM多特征融合分类模型识别率高达92.39%。基于PSO-SVM的多特征分类模型识别率高,可以及时、便捷、高效地识别黄花菜病害。     

Speedy Recognition of Noise in Real Time Signal Based on Wavelet Transform

It's necessary to process position and gesture of rocket on real time within the flight of rocket. This work should base exact data. But much noise has been found in the data and the noise recognize is regarded as a focus. A speedy online arithmetic recognizing noise is proposed based on wavelet transform. The computing complexity measured by time of this arithmetic is a constant which is greatly reduces the works of calculation of wavelet transform. It can recognize the noise fast when the signal is gathered. The applications in these problems show that the effective arithmetic satisfies the needs of real time and can handle the real time data measured in other yields.     

Detecting Technology of Cabel Surface Defect Based on DSP and Image Recognition

The damage of bridge cable directly hazards the normal use of bridge and the safety of people. The paper discussed the signal acquisition and detection system which takes embedded DSP processor as core technology, combined with CCD, video decode chip, SDRAM, FLASH ROM, and established the fast measuring algorithm of the surface corrosion and damage, which includes deteminating the processing region, improving medium filter, sobel edge detection and dynamic threshold segment, close operation of morphology, computing defect area and saving. The experimental resuits show that this method implemented the real-time, non-contact and intelligence detecting of cable surface defect and gives the specifications of the system.     

基于纹理特征和神经网络的图像识别

图像特征是用来区分图像内部最基本属性或特征的,它包括了几何特征、形状特征、颜色特征、纹理特征等,在这些特征中纹理特征理解过于复杂,至今尚未有一个大家都认同的定义,从而也就使得纹理分析变得很困难。该文主要研究图像的分类识别,采用基于灰度共生矩阵的纹理特征的提取。以8位灰度图为研究对象,依次对图像进行4个方向(0°、45°、90°、135°)的扫描,并分别建立灰度共生矩阵,从而提取纹理特征参数,然后建立BP神经网络模型,用提取的纹理特征值,来训练BP神经网络模型,使网络模型能够识别图像的内容。此文提出了用纹理特征训练神经网络的方法来识别图像,研究结果表明,纹理是图像分析中一个非常重要的特征,所提取的图像特征是描述图像相当有效的参数。     

基于国产基础软件的蔬菜病虫害识别防治系统的设计与实现

为提升蔬菜生产管理和蔬菜质量安全水平,建立方便、实用、低成本的蔬菜病虫害识别防治系统,本研究基于国产基础软件,搭建了基于中间件的系统运行支撑平台,采用拖拽式向导模式构件技术,构建了蔬菜病虫害识别防治系统,并在海南省进行了推广和应用。系统实现了关于蔬菜病虫害的知识浏览、智能诊断、农事指导三个核心功能,实现了基于国产基础软件运行环境对蔬菜病虫害识别防治的信息化管理。系统有效地帮助植保人员对蔬菜病虫害进行了识别防治,对增强蔬菜种植管理水平和提升农业生产信息化水平起到了一定的促进作用。     

基于深度卷积神经网络的储粮害虫图像识别

[目的]为了防治储粮害虫带来的危害,借助计算机对储粮害虫进行有效的图像识别是具有重要意义的。[方法]针对基于图像的储粮害虫多分类识别问题,引入了基于深度卷积神经网络的储粮害虫图像识别方法。该方法与传统的储粮害虫识别方法相比,大幅度简化了数据预处理过程,[结果]同时该方法在识别精确度方面达到了97.61%,也明显优于传统方法。[结论]因此基于深度卷积神经网络的储粮害虫识别方法具有较高的实用性,且具有进一步研究和推广的意义。     

基于迁移学习的园艺作物叶部病害识别及应用

为给农户提供质优价廉的园艺作物叶部病害识别服务,提出基于迁移学习的模型训练及基于Flask的Web部署方法.对PlantVillage数据集进行预处理,分别使用ResNet18、ResNet50和ResNet152 3种模型进行迁移学习训练,得到3种识别模型.利用Flask将模型部署到服务器上.3种识别模型对苹果等14...