Feature fusion method for edge detection of color images

A novel feature fusion method is proposed for the edge detection of color images. Except for the typical features used in edge detection, the color contrast similarity and the orientation consistency are also selected as the features. The four features are combined together as a parameter to detect the edges of color images. Experimental results show that the method can inhibit noisy edges and facilitate the detection for weak edges. It has a better performance than conventional methods in noisy environments.

A system for detection of cervical precancerous in field emission scanning electron microscope images using texture features

This study develops a novel cervical precancerous detection system by using texture analysis of field emission scanning electron microscopy(FE-SEM)images.The processing scheme adopted in the proposed system focused on two steps.The first step was to enhance cervical cell FE-SEM images in order to show the precancerous characterization indicator.A problem arises from the question of how to extract features which characterize cervical precancerous cells.For the first step,a preprocessing technique called intensity transformation and morphological operation(ITMO)algorithm used to enhance the quality of images was proposed.The algo-rithm consisted of contrast stretching and morphological opening operations.The second step was to characterize the cervical cells to three classes,namely normal,low grade intra-epithelial squamous lesion(LSIL),and high grade intra-epithelial squamous lesion(HSIL).To differen-tiate between normal and precancerous cells of the cervical cell FE-SEM images,human papillomavirus(HPV)contained in the surface of cells were used as indicators.In this paper,we investigated the use of texture as a tool in determining precancerous cell images based on the observation that cell images have a distinct visual texture.Gray level co-occurrences matrix(GLCM)technique was used to extract the texture features.To confirm the system's perfor-mance,the system was tested using 150 cervical cell FE-SEM images.The results showed that the accuracy,sensitivity and specificity of the proposed system are 95.7%,95.7%and 95.8%,respectively.

Computerized Detection of Limbal Stem Cell Deficiency from Digital Cornea Images

Limbal Stem Cell Deficiency(LSCD)is an eye disease that can cause corneal opacity and vascularization.In its advanced stage it can lead to a degree of visual impairment.It involves the changing in the semispherical shape of the cornea to a drooping shape to downwards direction.LSCD is hard to be diagnosed at early stages.The color and texture of the cornea surface can provide significant information about the cornea affected by LSCD.Parameters such as shape and texture are very crucial to differentiate normal from LSCD cornea.Although several medical approaches exist,most of them requires complicated procedure and medical devices.Therefore,in this paper,we pursued the development of a LSCD detection technique(LDT)utilizing image processing methods.Early diagnosis of LSCD is very crucial for physicians to arrange for effective treatment.In the proposed technique,we developed a method for LSCD detection utilizing frontal eye images.A dataset of 280 eye images of frontal and lateral LSCD and normal patients were used in this research.First,the cornea region of both frontal and lateral images is segmented,and the geometric features are extracted through the automated active contour model and the spline curve.While the texture features are extracted using the feature selection algorithm.The experimental results exhibited that the combined features of the geometric and texture will exhibit accuracy of 95.95%,sensitivity of 97.91% and specificity of 94.05% with the random forest classifier of n=40.As a result,this research developed a Limbal stem cell deficiency detection system utilizing features’fusion using image processing techniques for frontal and lateral digital images of the eyes.

Melanoma Detection Based on Hybridization of Extended Feature Space

Melanoma is a perfidious form of skin cancer.The study offers a hybrid framework for the automatic classification of melanoma.An Auto-matic Melanoma Detection System(AMDS)is used for identifying melanoma from the infected area of the skin image using image processing techniques.A larger number of pre-existing automatic melanoma detection systems are either commercial or their accuracy can be further improved.The research problem is to identify the best preprocessing technique,feature extractor,and classifier for melanoma detection using publically available MED-NODE data set.AMDS goes through four stages.The preprocessing stage is for noise removal;the segmentation stage is for extracting lesions from infected skin images;the feature extraction stage is for determining the features like asymmetry,border,and color,and the classification stage is to classify the lesion as benign or melanoma.The infected input image for the AMDS may contain impurities such as noise,illumination,artifacts,and hairs.In the proposed methodology an algorithm LePrePro is proposed for the prepro-cessing stage for denoising and brightness cum contrast normalization and another algorithm LeFET is proposed for extending the feature vector space in the feature extraction stage using a hybrid approach.In the study,a novel approach has been proposed in which different classifiers,feature extractions,and data preprocessing steps of the AMDS are compared.In a conclusion,this comparison revealed that on experimentation using Med-Node and ISIC 2017 Dataset,the best results included Gaussian blur as the best data preprocessing step,Extended feature vector which is the combination of Hue Saturation Value(HSV),and Local Binary Pattern(LBP)was the best feature extraction method,and the ensemble bagged tree was the best classification technique on the Med-Node data sets with 99%Area Under the Receiver Operating Characteristic Curve(AUC),93.52%accuracy,90.82%sensitivity,and 98.36%specificity in the proposed automatic melanoma detection system.

Study on Image Recognition Algorithm for Residual Snow and Ice on Photovoltaic Modules

The accumulation of snow and ice on PV modules can have a detrimental impact on power generation,leading to reduced efficiency for prolonged periods.Thus,it becomes imperative to develop an intelligent system capable of accurately assessing the extent of snow and ice coverage on PV modules.To address this issue,the article proposes an innovative ice and snow recognition algorithm that effectively segments the ice and snow areas within the collected images.Furthermore,the algorithm incorporates an analysis of the morphological characteristics of ice and snow coverage on PV modules,allowing for the establishment of a residual ice and snow recognition process.This process utilizes both the external ellipse method and the pixel statistical method to refine the identification process.The effectiveness of the proposed algorithm is validated through extensive testing with isolated and continuous snow area pictures.The results demonstrate the algorithm’s accuracy and reliability in identifying and quantifying residual snow and ice on PV modules.In conclusion,this research presents a valuable method for accurately detecting and quantifying snow and ice coverage on PV modules.This breakthrough is of utmost significance for PV power plants,as it enables predictions of power generation efficiency and facilitates efficient PV maintenance during the challenging winter conditions characterized by snow and ice.By proactively managing snow and ice coverage,PV power plants can optimize energy production and minimize downtime,ensuring a sustainable and reliable renewable energy supply.

一种基于航拍红外图像的光伏热斑故障分类检测方法

针对航拍光伏红外图像热斑检测方法中小目标特征易丢失问题,提出一种光伏热斑故障分类检测方法。首先将多头自注意力机制结合CSPNet结构进行改进,提出CSPMAT网络,再将其引入New CSP-Darknet网络,构建CSPMAT-Darknet模型,实现了光伏组件热斑定位及分类。实验结果表明:该模型在小目标检测任务中的性能显著提升,且在目标尺寸差异较大的故障分类检测任务中,均值平均精度达到82.92%,提高了13.97个百分点,具有良好的检测精度和泛化能力。

基于轻量化YOLOv5s的光伏热斑检测定位方法

针对目前目标检测技术在检测光伏热斑效应时模型检测速度低、计算复杂、模型结构复杂等问题,提出基于轻量化YOLOv5s的光伏热斑检测定位方法。首先,以YOLOv5s为基础模型,引入轻量网络ShuffleNetV2改进YOLOv5s的主干网络,利用其分组卷积和通道混洗的设计思想,减少模型参数和计算量,同时保持较高的准确率。其次,引入轻量级卷积GSConv优化YOLOv5s的Neck部分,利用其深度可分离卷积结合标准卷积的形式,降低计算复杂度,优化整体模型。最后利用数据集进行验证。结果表明,轻量化后的模型在保证较高精度的前提下,能够提高检测速度、减少参数量和计算量。

基于YOLOv5改进的红外目标检测算法

为了解决红外图像特征少、对比度不佳导致目标检测时精度低的问题,采用增加一个额外的预测特征层的方法,以提高原始YOLOv5在红外图像中的识别率;通过添加坐标注意力机制,优化红外目标强特征提取,提升检测准确度;再使用双向特征金字塔网络优化特征融合,增强模型表达能力,降低冗余计算;最后解决检测定位差和边界框回归任务中样本不平衡,采用focal-EIOU作为模型的边界框损失函数,提高收敛速度,并专注于高质量的锚框回归。结果表明,改进的YOLOv5在FLIR数据集上的准确率达到了85.3%,相比于原始网络模型提高了4.2%,具有较高的检测准确率。这一结果为在嵌入式设备上部署该软件提供了可行性。

一种基于改进RefineDet的管道数字射线成像缺陷图像检测方法

为提高管道缺陷图像检测的准确率,提出了一种基于改进RefineDet的管道数字射线成像(digital radiography, DR)缺陷图像检测模型。该模型针对管道DR缺陷图像数据少、目标少等特点,从以下三个方面进行改进。首先,在骨干网络设计方面,使用Swin transformer代替VGG16作为主干网络,在提高特征提取能力的同时减少主干网络参数量。其次,针对管道DR缺陷图像目标数量较少而易受背景干扰问题,通过在主干网络与特征融合阶段之间加入全局注意力模块来强化模型对重要特征的关注,从而提高检测性能。最后,在后处理阶段,针对传统的非最大值抑制算法直接去除非最好预测框问题,使用软非最大值抑制算法以更合理的方式去除非最优预测框。结果表明:该方法能够有效实现管道DR缺陷图像的检测,并且相比于其他4种常用的目标检测模型,提出的模型可以有效提升管道DR缺陷图像检测的准确率,研究结果可为DR缺陷图像检测提供技术支撑。

同步码字优化降噪的声纳图像多目标检测方法

针对海底侧扫声纳图像分辨率低、噪声污染严重导致水下目标检测不准确的问题,提出一种结合同步码字优化降噪的水下声纳图像目标检测方法。利用同步码字优化对声纳图像中的乘性噪声进行降噪处理,从而使图像中的水下目标物获得更好的视觉与检测效果,同时对声纳图像进行相应的数据集扩充。最后利用适合本文方法的YOLO系列中的YOLOv7对降噪后声纳图像中的目标物体进行检测,并在其特征网络中加入了卷积块注意模块,从而加强对目标的特征提取。仿真结果分析得出,同步码字优化降噪与YOLOv7相结合的目标检测方法,可使目标置信度达到79%,相较于降噪前的目标检测置信度提高16%,对于目标较小的物体,能更好地改善漏检与误检情况。

基于改进的CenterNet变电站设备红外温度检测方法

红外检测能够检测变电站电力设备温度异常,降低安全事故发生的概率,因此,提出一种基于改进的CenterNet目标检测算法模型CenterNet_PRO;该算法采用了ShuffleNet V1/V2作为骨干网络、引入了FPN来提取多尺度特征,为了克服不同尺度目标检测的难点、增加旋转角度回归分支,用于预测目标的旋转角度以及改进的IoU Loss进行优化,进一步提高模型检测速度和准确率;通过阈值分割法提取电力设备表面温度并分析计算,设计制定电力设备温度缺陷判断规范、温度警告阈值,根据该规范即可判断电力设备的相关缺陷;实验结果表明,改进的CenterNet模型平均精度达到了90%,相比于传统的CenterNet模型,平均精度提高了1.3个百分点,可以满足实际变电站场景下对电力设备红外检测的高要求。

基于改进细化算法的裂纹检测研究

为了提高管道裂纹缺陷的检测水平,提出一种基于改进细化算法的管道裂纹检测算法,并对其有效性进行研究.首先,利用管道机器人搭载的视觉传感器获取裂纹图像,并对图像进行预处理;然后,提出基于改进的OTSU算法和骨架的长短轴算法分别进行图像分割和特征提取,计算出目标的长度、宽度和面积等几何特征;最后,对特征进行分析并检测裂纹目标.实验结果表明,与传统算法相比,本文所提算法准确率较高,能够有效计算裂纹的几何特征值并进行检测,满足管道裂纹检测的要求.

基于机器视觉的植球缺陷识别技术

植球缺陷是芯片制造过程中常见的问题之一,如果不及时检测和修复,可能会导致芯片性能下降或故障。基于机器视觉技术,研究并设计了一种用于芯片植球缺陷检测的方法,旨在提高芯片制造过程中的质量控制能力。通过收集大量包含正常与缺陷样本的芯片植球图像数据集,运用计算机视觉算法进行深入处理与分析。研究过程中,对图像进行了预处理,例如去噪和图像增强,以提升缺陷检测的准确性。进一步采用图像分割算法,有效地将芯片植球从图像背景中分离,便于精确检测缺陷。此外,利用特征提取算法识别芯片植球的独特性质,并通过分类算法准确区分正常与缺陷芯片样本。不仅提高了芯片缺陷检测的精准度,而且为工业生产中的质量控制提供了重要依据,具有积极的现实意义。

基于计算机视觉的稻谷霉变程度检测

为了实现无损检测稻谷储藏中的霉变,该研究以引起稻谷霉变的5种常见真菌(米曲霉、黑曲霉、构巢曲霉、桔青霉和杂色曲霉)为对象,首先进行真菌培养,制成悬浮液,然后将悬浮液接种到稻谷样品中,对稻谷样品模拟储藏,确定不同霉变程度的稻谷类型,划分为对照组(无霉变)、轻微霉变组和严重霉变组。利用计算机视觉系统对三组稻谷样品进行图像采集和图像处理,提取灰度、颜色和纹理特征,共获取68个图像特征。采用支持向量机(support vector machines,SVM)和偏最小二乘法判别分析(partial least squares discriminant analysis,PLS-DA)构建模型,分别用于无霉变稻谷与霉变稻谷的区分和稻谷霉变类型区分。为了降低模型复杂度和数据冗余,利用连续投影算法(successive projections algorithm,SPA)来消除原始数据变量间的共线性,优选特征值。结果表明:利用所有参数构建的SVM模型能够很好的区分对照组与霉变组,其中建模集和验证集总体区分准确率分别为99.7%和98.4%;SVM模型对于稻谷严重霉变类型的区分效果要优于轻微霉变稻谷,其中对稻谷轻微霉变类型建模集和验证集总体区分的准确率分别为99.3%和92.0%,对稻谷严重霉变类型区分的总体准确率分别为100%和94%,且整体上SVM模型的效果要优于PLS-DA模型。而基于SPA优选特征构建的模型区分结果表明,SVM模型区分效果优于PLS-DA模型,其中,在建模集和验证集中,对无霉变和霉变稻谷总体区分准确率分别为99.8%和99.5%,对稻谷轻微霉变种类区分总体准确率分别为99.8%和90.5%,对稻谷严重霉变种类区分总体准确率分别为100%和95.0%。因此,基于计算机视觉对稻谷霉变检测是可行的,而且SPA优选特征能够较好反映稻谷霉变特征,基于优选特征和SVM模型能够较好地稻谷霉变进行识别和区分,结果较好,可以为实际应用提供技术支持和参考。

基于特征的遥感图像自动配准算法

面向光学图像的多时相、多光谱、多传感器图像的自动配准 ,本文描述一个基于特征的高精度图像配准算法 .它以点映射配准技术为基础 ,处理具有全局仿射几何失真的图像配准问题 .首先 ,通过边缘检测和相应的后处理提取封闭边界 ;其次 ,根据边界链码相关和区域不变矩匹配策略建立边界的对应 ,并对对应重心即匹配点对进行一致性检测获得基元控制点 ;最后 ,估计初始变换参数 ,并通过显著点片的相关匹配来增加控制点个数 ,迭代修正变换参数以提高配准精度 .多种遥感图像数据的配准实验和对比试验证实了的自动算法具有较高的可靠性和配准精度 .

基于小波变换的复杂航空图像的边缘提取

提出了一种基于小波变换的复杂航空照片的目标边缘自动提取的新方法。该方法利用了小波变换,因而具有良好的噪声抑制能力及完备的边缘保持能力,在计算机自动判读系统的目标边缘提取中有效可行。

一种全天候驾驶员疲劳检测方法研究

针对驾驶员疲劳状态检测问题,本文提出了基于红外图像处理和生理特征—心率的全天候疲劳检测算法,采用模糊神经网络专家系统对驾驶员的疲劳状态识别。假设驾驶员驾驶汽车的初始阶段(前十分钟内)是清醒的,这样在前十分钟内,模糊神经网络处于学习阶段并记忆驾驶员的状态,在十分钟之后模糊神经网络处于离线自学习,在线对驾驶员状态实时识别。通过实验表明该检测方法克服了光线和气候的影响,该识别方法具有较强的自适应能力。

一种用于掌纹识别的线特征表示和匹配方法(英文)

作为一种较新的生物特征,掌纹可用来进行人的身份识别.在用于身份识别的诸多特征中,掌纹线,包括主线和皱褶,是最重要的特征之一.本文为掌纹识别提出一种有效的掌纹线特征的表示和匹配方法,该方法定义了一个矢量来表示一个掌纹上的线特征,该矢量称为线特征矢量(line feature vector,简称 LFV).线特征矢量是用掌纹线上各点的梯度大小和方向来构造的.该矢量不但含有掌纹线的结构信息,而且还含有这些线的强度信息,因而,线特征矢量不但能区分具有不同线结构的掌纹,同时也能区分那些具有相似的线结构但各线强度分布不同的掌纹.在掌纹匹配阶段,用互相关系数来衡量不同线特征矢量的相似性.实验表明,LFV方法无论是在速度、精度,还是在存储量方面都能满足联机生物识别的要求.

路况视频能见度检测算法

针对现有气象能见度仪价格昂贵且无法全程分析的不足,提出一种利用路况视频图像的能见度检测算法,通过目标物预置获得距离信息,通过图像预处理得到含有目标物的去噪图像。利用目标物的相互关系和SAD图像匹配算法从图像上分割出目标物,利用Kohler的一致对比度法实现目标物图像的人眼模拟,进而对其特征进行拟合,得出能见度值,实现基于路况视频的能见度检测。实验结果表明,该算法与人眼观察一致,对噪声有较好的鲁棒性。

基于机器视觉的光栅表面缺陷检测系统

为了实现对光栅表面缺陷的自动检测,开发了一种基于机器视觉的光栅表面缺陷检测系统。首先,搭建了系统硬件平台,分析并解决了关键检测难点;然后,在分析比较了几种传统边缘检测算法的基础上,介绍了边缘重绘算法,对光栅表面图像进行边缘检测;最后,根据各种光栅表面缺陷特征,选取多个阈值定义多种光栅表面缺陷,提取特征缺陷并标记,开发了检测算法。实验结果表明:该系统检测精度达微米级,单个光栅检测效率为30 s左右,缺陷识别率在94%以上,误检率在4%以下,可在线实时检测多种光栅表面缺陷。