基于SVM的离线图像目标分类算法

目标分类是计算机视觉与模式识别领域的关键环节. SVM(支持向量机)是在统计学习理论基础上提出的一种新的机器学习方法.提出一种支持向量机结合梯度直方图特征的离线图像目标分类算法.首先对训练集进行预处理,然后对处理后的图片进行梯度直方图特征提取,最后通过训练得到可以检测图像目标的分类器.利用得到的分类器对测试图片进行测试,测试结果表明,对目标分类检测有良好的效果.     

车牌超分辨率重建与识别

为了从图片中快速准确地识别车牌,提出一种结合图像超分辨率技术的车牌识别方案。车牌图片具有明显的特定的模式特征,只是具体的字符编码不同。因此车牌图片非常适合做超分辨率重建。本文提出的系统主要由车牌检测定位、车牌超分辨率重建、字符分割、字符识别等模块组成。综合基于边缘、基于颜色和基于最大稳定极值区域三种车牌检测策略并采用并行编程方法来综合检测结果得到候选车牌。采用车牌图片正负样本来训练支持向量机分类器。得到分类器模型后对候选车牌判决得到真正的车牌。随后对真实车牌图片进行超分辨率重建。该部分主要由基于固定邻域回归的方法实现。这种方法综合了稀疏字典学习和领域嵌入的方法,比较好的兼顾了准确率和计算速度。运用OpenCV提供的图像处理库来对重建后的图片做字符分割。得到单独的字符图片后采用人工神经网络进行识别。识别前先使用一定数量的字符图片对网络进行有监督训练获取识别模型。采用一个单隐层的神经网络,运用反向传播算法进行训练得到识别模型。最后提取字符图片的特征并输入网络进行分类完成识别。为了测试系统的表现,在实际场景中采集了一百张车牌图片作为测试集。实验表明,该系统具有较高识别准确率和较快的处理速度。     

基于SVM-KNN的半监督托攻击检测方法

针对支持向量机方法在标记用户数据不充分的情况下无法有效实现托攻击检测的不足,提出一种基于SVM-KNN的半监督托攻击检测方法。根据少量标记用户数据训练一个初始SVM分类器,利用初始SVM对大量未标记用户数据进行分类,挑选出分类边界附近有可能成为支持向量的样本点,利用KNN分类器优化边界向量的标记质量,再将重新标注过的边界向量融入训练集,迭代训练逐步改善SVM的分类边界,最终获得系统决策函数。实验结果表明在标记用户数据较少的情况下,方法能有效提高托攻击的检测精度和效率,具有较强的推广能力。     

基于Adaboost算法的多姿态人脸实时视频检测

针对实时视频中的多姿态人脸检测问题,应用扩展的类Haar特征,训练能有效检测多种姿态和多种旋转角度人脸的分类器;并使用该分类器实现了一个实时视频的多姿态人脸检测系统.该系统分为训练和检测两个子系统,训练系统应用大量包含正反例子的图片进行训练,得到分类器;检测系统首先使用DiectShow从USB摄像头获取图像,然后读入分类器,对图像进行检测并显示.实验结果表明,该系统能够快速准确地在视频中检测出多种姿态的人脸,有较强的实用价值.     

基于广义霍夫变换的室外场景行人检测研究

提出一种基于广义霍夫变换的室外场景行人检测方法.首先从少量标注图片中随机地提取行人图像碎片构造碎片字典,然后使用图像碎片对每一幅训练图片计算特征向量.为了能够在静态图片中快速地检测行人,使用Gentleboost算法训练检测器,在每一次迭代时学习一个决策树桩弱分类器,该弱分类器可以从高维特征向量中选择一个当前区分度最好的碎片特征.在运行检测器时,所有的弱分类器在测试图片中对于行人的可能出现位置进行投票.最后,将各个弱分类器的投票结果进行叠加,并用设定的检测阈值剔除得分较低的检测结果后得到检测输出.在LabelMe数据集上的实验表明,该方法可以快速地在静态图片中检测出行人,需要较少的训练数据且有效地解决了部分遮挡问题.     

深度学习在仿生眼监控系统中的应用

在视频监控系统中,由于受到复杂的背景、环境光线变化以及设备本身性能的限制,导致目标检测算法设计难度的加大,而传统的目标检测算法通常依赖于人工选择特征,难以从海量的数据中自动学习得到一个有效的分类器.基于深度学习算法,构建了一个卷积神经网络,并利用仿生眼视频监控系统中采集的人、车图像进行训练,在此基础上设计若干实验对深度学习网络特性进行分析,证明了训练集中各个类别样本的分布以及小样本训练的情况下对深度学习的训练结果会造成较大的影响.     

基于深度学习的微信小程序图像识别系统

为提高用户在移动端的图像识别体验,设计基于深度学习的微信小程序图像识别系统。该系统的总体架构包括微信小程序前端和后台服务器两部分。在微信小程序端,用户可通过相机拍摄或加载本地图片进行图像采集。在服务器端,经过图像预处理和基于深度学习的目标识别,识别结果再通过应用程序编程接口（Application Programming Interface,API）传输回微信小程序进行展示。选择CIFAR-100数据集,设计并训练卷积神经网络用于目标识别,最终的测试结果证明了该系统的有效性。     

用于提高谷歌图像搜索结果的二分类器在线学习方法

对于基于关键词的图像检索,利用检索结果的视觉相似性学习二分类器有望成为改善检索结果的最有效途径之一. 为改善搜索引擎的搜索结果,本文提出一种算法框架并且基于此框架着重研究训练数据选择这一关键问题. 训练数据选择过程由两个阶段组成:1）训练数据初始化以开始分类器学习过程;2）分类器迭代学习过程中的动态数据选择. 对于初始训练数据的选择,我们探讨了基于聚类和基于排序两种方法,并且对比了自动训练数据选择与人工标注的结果. 对于动态数据选择,我们比较了支持向量机和基于最大最小后验伪概率的贝叶斯分类器的分类效果. 组合上述两个阶段的不同方法,我们得到了8种不同的算法,并将其用于谷歌搜索引擎进行基于关键词的图像检索. 实验结果证明,如何从含有噪声的搜索结果中选择训练数据是搜索结果改善的关键问题. 实验显示我们的方法能够有效的改善谷歌搜索的结果,尤其是排序在前的结果. 尽早为用户提供更相关的结果能够更大程度的减少用户逐个翻页查看结果的工作. 另外,如何使自动训练数据选择与人工标注媲美仍是需要继续研究的一个问题.     

基于联邦卷积神经网络的鱼类检测系统

由于鱼类数据的多样性以及应用的广泛性,为了进一步提高鱼类检测的效率,以及在处理鱼类图片时提取到更高维的特征来提高鱼类检测的准确率,将卷积神经网络与联邦学习相结合,将鱼类图片数据按照非独立同分布的形式分发给用户。用户在本地训练模型,并将训练好的模型参数上传到云端,云端将完成模型参数的聚合与更新,并将更新好的参数返回到用户的终端,各个用户开始下一轮训练。以此过程来训练网络,并模拟联邦学习的过程。最后,用联邦卷积神经网络、联邦学习以及卷积神经网络分别对野生鱼类数据集上鱼类图片进行图像检测与识别,并将结果做对比。结果表明,联邦卷积神经网络模型最终的分类准确率为33.3%,传统的联邦学习的准确率为26.67%,Resnet50的准确率为87.97%,可以看出联邦卷积神经网络的分类准确率远高于传统的联邦学习。并且联邦卷积神经网络模型在训练轮数较少的情况下就可以得到较好的实验结果。联邦学习作为分布式计算的重要组成部分,它提供的快速模糊化处理以及数据独立的特性,为鱼类分类的效率和数据保护提供了有力保障。卷积神经网络也提高了联邦学习的学习效率。这使得提出的联邦卷积神经网络分类系统相比于传统的联邦学习在分...     

基于光流法和卷积神经网络的跌倒行为检测方法

基于光流法和卷积神经网络,提出一种室内跌倒行为检测方法.在数据预处理方面,使用光流法把监控视频转化为由光流图像组成的动作序列;在模型方面,采用卷积神经网络VGG-16对输入动作序列进行训练和优化,根据softmax分类器输出结果不断调整权重和偏量;在训练过程中,制作了跌倒数据集并采用迁移学习训练策略解决训练过程中跌倒行...     

Feel the image: The role of emotions in the image-seeking process

Images are characterized by a complex system of attributes. One of the most elusive properties of an image is its emotiveness – the ability to trigger an emotional reaction in the viewer. Prior research has demonstrated that users react to image emotiveness, yet to date there is little theoretical understanding and scarce empirical evidence as to the role played by emotiveness in the image retrieval process. Our study aims to fill this gap and investigates how users perceive the emotional content of the image-seeking task; explicate this emotional content in search keywords; and react to image emotiveness in selecting relevant images. Using a multimethod approach that combines quantitative and qualitative analyses, we performed three experiments where participants searched for images, both within the predefined set of images and in an open Web environment. Our findings suggest that although seekers rarely explicate image emotiveness in search keywords, their decision to select relevant images is largely dependent on their perception of the emotional content of the task and images. We discuss implications for research on emotions in image retrieval, as well as practical implications for designers of search systems.     

Product appearance detection based on visual keywords matching

The product appearance detection device based on the USB (Universal Serial Bus) camera is studied. The USB camera is used as the front device of the image acquisition and the host computer is a PC for the detection system. Aiming at low resolution of USB camera, the super-resolution image reconstruction method based on sparse representation is used to overcome the low resolution for the low-cost imaging sensor and improve the image resolution. A method of product appearance detection based on visual keywords is proposed. Visual keywords are used to describe characteristics of defects. Visual keyword extraction is performed on sample images and test images, and visual keywords are used for matching. Identify defective products by setting different thresholds. Experiments show that the proposed system has better detection results and offers good robustness to the angle and illumination of the image from the product appearance. In this paper, the theory of visual keywords is applied to product appearance detection for the first time, which provides a reference for further research.     

Geotag propagation in social networks based on user trust model

In the past few years sharing photos within social networks has become very popular. In order to make these huge collections easier to explore, images are usually tagged with representative keywords such as persons, events, objects, and locations. In order to speed up the time consuming tag annotation process, tags can be propagated based on the similarity between image content and context. In this paper, we present a system for efficient geotag propagation based on a combination of object duplicate detection and user trust modeling. The geotags are propagated by training a graph based object model for each of the landmarks on a small tagged image set and finding its duplicates within a large untagged image set. Based on the established correspondences between these two image sets and the reliability of the user, tags are propagated from the tagged to the untagged images. The user trust modeling reduces the risk of propagating wrong tags caused by spamming or faulty annotation. The effectiveness of the proposed method is demonstrated through a set of experiments on an image database containing various landmarks.     

图像中物体空间关系的表示及在检索中的应用

基于图像中物体之间的空间关系的图像检索往往受困于待处理的图像中物体种类和空间位置难以自动准确地获取。文中基于物体识别算法的输出,提出一种对物体空间关系的三元组表示法,给出基于这种表示方法对图像索引、相似度计算和检索排序的方法及允许用户使用查询词和空间关系表达查询需求的二维输入界面,并实现原型系统。这种表示法具有良好的鲁棒性,可容忍物体识别算法一定程度的误差,将物体识别得到的置信度加入三元组表示法置信度计算和排序算法中,减少物体识别结果误差对检索性能的影响。在原型系统上的实验表明,该系统在实验中对包含物体位置关系的检索给出更准确的结果,在NDCG@m、MAP、F@m上均优于现有系统。     

航空遥感图像深度学习目标检测技术研究进展

航空遥感图像目标检测旨在定位和识别遥感图像中感兴趣的目标,是航空遥感图像智能解译的关键技术,在情报侦察、灾害救援和资源勘探等领域具有重要应用价值。然而由于航空遥感图像具有尺寸大、目标小且密集、目标呈任意角度分布、目标易被遮挡、目标类别不均衡以及背景复杂等诸多特点,航空遥感图像目标检测目前仍然是极具挑战的任务。基于深度卷积神经网络的航空遥感图像目标检测方法因具有精度高、处理速度快等优点,受到了越来越多的关注。为推进基于深度学习的航空遥感图像目标检测技术的发展,本文对当前主流遥感图像目标检测方法,特别是2020—2022年提出的检测方法,进行了系统梳理和总结。首先梳理了基于深度学习目标检测方法的研究发展演化过程,然后对基于卷积神经网络和基于Transformer目标检测方法中的代表性算法进行分析总结,再后针对不同遥感图象应用场景的改进方法思路进行归纳,分析了典型算法的思路和特点,介绍了现有的公开航空遥感图像目标检测数据集,给出了典型算法的实验比较结果,最后给出现阶段航空遥感图像目标检测研究中所存在的问题,并对未来研究及发展趋势进行了展望。     

基于多特征融合的图像主体显著性判断

从视觉角度来说,视觉显著性图像是指主体突出的图像,比起内容散乱的图像,此类图像往往更能吸引用户的关注,也更符合用户对图片检索的使用需求。提出了一种图像主体视觉显著性判断方法,采用“中心围绕”计算原则在多特征融合的基础上应用支持向量机训练,建立了一个分类模型,并且可以给出表征图像显著程度的得分。实验表明,该模型有较高的分类正确率,并且将该模型应用于图像检索重排序、图像上传自动审核等应用时,可以得到更接近人工操作的结果,降低人力资源成本。     

Image Indexing and Retrieval Using Object-based Point Feature Maps

Multimedia data such as audios, images, and videos are semantically richer than standard alphanumeric data. Because of the nature of images as combinations of objects, content-based image retrieval should allow users to query by image objects with finer granularity than a whole image. In this paper, we address a web-based object-based image retrieval (OBIR) system . Its prototype implementation particularly explores image indexing and retrieval using object-based point feature maps. An important contribution of this work is its ability to allow a user to easily incorporate both low- and high-level semantics into an image query. This is accomplished through the inclusion of the spatial distribution of point-based image object features, the spatial distribution of the image objects themselves, and image object class identifiers. We introduce a generic image model, give our ideas on how to represent the low- and high-level semantics of an image object, discuss our notion of image object similarity, and define four types of image queries supported by the OBIR system. We also propose an application of our approach to neurological surgery training.     

融合Snake模型与拓扑路线的图像分割算法

为了能够提高图像边缘检测的准确度,提出一种新型图像处理算法.该算法是基于主动轮廓方法和拓扑路线相结合的方法,目的是提高图像检测过程的精确度.该算法提出了新型技术来整合拓扑路线和主动轮廓方法各自的优点.将基于拓扑路线的初始分割边界作为Snake模型输入信号,并逐步演化成为最终对象的分割边界.实验结果表明,该算法可以处理低对比度图像,同时可以提高针对弱图像边界进行分割的准确度,取得了更好的图像分割和边缘检测效果,说明该算法有改进低对比度和自动图像分割系统的处理能力.     

A content-based goods image recommendation system

The information of e-commerce images varies and different users may focus on different contents of the same image for different purpose. So the research on recommendation by computers is becoming more and more important. But retrieval based only on keywords obviously falls short for massive numbers of resource images. In this paper, we focus on a recommendation system of goods images based on image content. Goods images have a relatively homogenous background and have a wide range of applications. The recommendation consists of three stages. First, the image is pre-processed by removing the background. Second, a weighted representation model is proposed to represent the image. The separated features are extracted and normalized, and then the weights of each feature are computed based on the samples browsed by the users. Third, a feature indexing scheme is put forward based on the proposed representation. A binary-tree is used for the indexing, and a binary-tree updating algorithm is also given. Finally, the recommended images are given by a features combination searching scheme. Experimental results on a real goods image database show that our algorithm can achieve high accuracy in recommending similar goods images with high speed.     

Image tag recommendation based on friendships

The tagging systems have been studied by many researchers in the past decade. Tagging methods have been widely used on the web for searching and recommending images. Social tags are the keywords annotated by users to the images, which contains the information for searching and classifying the images. Tag recommendation system allows mitigating the individual preferences to annotate and recommender images. However, irrelevant and noise tags are frequently included in tags. In this paper, we propose image tag recommendation based on the friends’ relationships in social network (TRboFS) to recommender tags for a new image, both the tags assigned to the favorite images and the friendships of the users who upload the image are employed to predict the tags of the images. Empirical analyses on real datasets show that the proposed approach achieves superior performance to existing approaches.