Facial expression feature extraction method based on improved LBP

Local binary pattern(LBP)is an important method for texture feature extraction of facial expression.However,it also has the shortcomings of high dimension,slow feature extraction and noeffective local or global features extracted.To solve these problems,a facial expression feature extraction method is proposed based on improved LBP.Firstly,LBP is converted into double local binary pattern(DLBP).Then by combining Taylor expansion(TE)with DLBP,DLBP-TE algorithm is obtained.Finally,the DLBP-TE algorithm combined with extreme learning machine(ELM)is applied in seven kinds of ficial expression images and the corresponding experiments are carried out in Japanese adult female facial expression(JAFFE)database.The results show that the proposed method can significantly improve facial expression recognition rate.

Binocular Vision Positioning Method for Safety Monitoring of Solitary Elderly

In nowadays society,the safety of the elderly population is becoming a pressing concern,especially for those who live alone.There might be daily risks such as accidental falling or treatment attack on them.Aiming at these problems,indoor positioning could be a critical way to monitor their states.With the rapidly development of the imaging techniques,wearable and portable cameras are very popular,which could be set on human individual.And in view of the advantages of the visual positioning,the authors propose a binocular visual positioning algorithm to real-timely locate the elderly indoor.In this paper,the imaging model has been established with the corrected image data from the binocular camera;then feature extraction has been completed to provide reference to adjacent image matching based on the binary robust independent elementary feature(BRIEF)descriptor,finally the camera movement and the states of the elderly have been estimated to distinguish their falling risk.In the experiments,the real-sense D435i sensors were adopted as the binocular cameras to obtain indoor images,and three experimental scenarios have been carried out to test the proposed method.The results show that the proposed algorithm can effectively locate the elderly indoor and improve the real-time monitoring capability.

Binocular Visual Navigation and Obstacle Avoidance of Mobile Robots Based on Speeded-Up Robust Features

This article presents a good robust and real-time system scheme of the mobile robot obstacle detection and navigation, which principle of work is based on the feature descriptor SURF. In this scheme, firstly, the image information of the mobile robot path was captured by the binocular camera; then the feature points were extracted and corresponding matched using SURF to the binocular images as the undetected obstacles; finally fixed the position of the objective by the parallax between the matching points combining with the binocular vision calibration model. Theoretical derivation and experimental results show that this scheme is more accurate for the detection and navigation of the interest points. It has fast matching speed and high accuracy and low error. So, it has certain practical effect and popularizing value for the mobile robot real-time obstacle avoidance and navigation.

细粒度图像分类上Vision Transformer的发展综述

细粒度图像分类(fine-grained image classification,FGIC)一直是计算机视觉领域中的重要问题。与传统图像分类任务相比,FGIC的挑战在于类间对象极其相似,使任务难度进一步增加。随着深度学习的发展,Vision Transformer(ViT)模型在视觉领域掀起热潮,并被引入到FGIC任务中。介绍了FGIC任务所面临的挑战,分析了ViT模型及其特性。主要根据模型结构全面综述了基于ViT的FGIC算法,包括特征提取、特征关系构建、特征注意和特征增强四方面内容,对每种算法进行了总结,并分析了它们的优缺点。通过对不同ViT模型在相同公用数据集上进行模型性能比较,以验证它们在FGIC任务上的有效性。最后指出了目前研究的不足,并提出未来研究方向,以进一步探索ViT在FGIC中的潜力。

Application of Computer Vision Technique to Maize Variety Identification

Variety identification is important for maize breeding, processing and trade. The computer vision technique has been widely applied to maize variety identification. In this paper, computer vision technique has been summarized from the following technical aspects including image acquisition, image processing, characteristic parameter extraction, pattern recognition and programming softwares. In addition, the existing problems during the application of this technique to maize variety identification have also been analyzed and its development tendency is forecasted.

基于EVision的双目视觉系统的设计与实现

设计和开发了一个排爆机器人的双目立体视觉系统。该系统使用Matlab7.0和机器视觉软件EVision6.2的EasyMultiCam进行图像捕获并和EasyMath模式匹配库进行图像特征匹配,能实时捕获排爆机器人周围的图像信息、进行摄像机标定、图像预处理、立体图像匹配、求取可疑目标物的特征点的立体坐标,并把图中实时显示在控制台上,控制排爆机器人准确地抓取可疑目标物。该机器人视觉系统成功的抓取实验,表明了它在精度上能够满足排爆机器人的项目要求。

Image Retrieval Based on Vision Transformer and Masked Learning

Deep convolutional neural networks(DCNNs)are widely used in content-based image retrieval(CBIR)because of the advantages in image feature extraction.However,the training of deep neural networks requires a large number of labeled data,which limits the application.Self-supervised learning is a more general approach in unlabeled scenarios.A method of fine-tuning feature extraction networks based on masked learning is proposed.Masked autoencoders(MAE)are used in the fine-tune vision transformer(ViT)model.In addition,the scheme of extracting image descriptors is discussed.The encoder of the MAE uses the ViT to extract global features and performs self-supervised fine-tuning by reconstructing masked area pixels.The method works well on category-level image retrieval datasets with marked improvements in instance-level datasets.For the instance-level datasets Oxford5k and Paris6k,the retrieval accuracy of the base model is improved by 7%and 17%compared to that of the original model,respectively.

Accurate vision measurement for kinematic parameters of satellite separation tests

The accurate measurement of kinematic parameters in satellite separation tests has great significance in evaluating separation performance. A novel study is made on the measuring accuracy of monocular and binocular, which are the two main vision measurement methods used for kinematic parameters. As satellite separation process is transient and high-dynamic, it will bring more extraction errors to the binocular. Based on the design approach of intersection measure and variance ratio, the monocular method reflects higher precision, simpler structure and easier calibration for level satellite separation. In ground separation tests, a high-speed monocular system is developed to gain and analyze twelve kinematic parameters of a small satellite. Research shows that this monocular method can be widely applied for its high precision, with position accuracy of 0.5 mm, speed accuracy of 5 mm/s, and angular velocity accuracy of 1 (°)/s.

Laser vision based adaptive fill control system for TIG welding

The variation of joint groove size during tungsten inert gas （TIG） welding will result in the non-uniform fill of deposited metal. To solve this problem, an adaptive fill control system was developed based on laser vision sensing. The system hardware consists of a modular development kit （MDK） as the real-time image capturing system, a computer as the controller, a D/A conversion card as the interface of controlled variable output, and a DC TIG welding system as the controlled device. The system software is developed and the developed feature extraction algorithm and control strategy are of good accuracy and robustness. Experimental results show that the system can implement adaptive fill of melting metal with high stability, reliability and accuracy. The groove is filled well and the quality of the weld formation satisfies the relevant industry criteria.

基于Vision Transformer的中文唇语识别

唇语识别作为一种将唇读视频转换为文本的多模态任务,旨在理解说话者在无声情况下表达的意思.目前唇语识别主要利用卷积神经网络提取唇部视觉特征,捕获短距离像素关系,难以区分相似发音字符的唇形.为了捕获视频图像中唇部区域像素之间的长距离关系,文中提出基于Vision Transformer(ViT)的端到端中文句子级唇语识别模型,融合ViT和门控循环单元(Gate Recurrent Unit,GRU),提高对嘴唇视频的视觉时空特征提取能力.具体地,首先使用ViT的自注意力模块提取嘴唇图像的全局空间特征,再通过GRU对帧序列时序建模,最后使用基于注意力机制的级联序列到序列模型实现对拼音和汉字语句的预测.在中文唇语识别数据集CMLR上的实验表明,文中模型的汉字错误率较低.

基于Vision Transformer的光伏组件红外图像故障检测

太阳能光伏板受制造、运输、安装以及环境因素的影响,易发生故障和损坏,造成能量损失。通过对电池板进行红外图像检测,可以估计电力生产的损失,降低运行和维护的成本。基于此,设计了一种基于Vision Transformer的光伏异常红外图像检测的方法,通过对异常红外图像的检测,达到对不同的故障类型进行分类的目的。Vision Transformer首先将输入进来的图片,每隔一定的区域大小划分图片块,然后将划分后的图片块组合成序列,并将组合后的结果传入Transformer特有的Multi-head Self-attention进行特征提取,最后利用Cls Token进行分类。实验结果表明基于本文方法的红外图像检测准确率可达到95.787%,高于Xception模型11.9%、高于VGG16模型17.74%。

基于单激光束信息的掘锚装备视觉定位方法研究

煤矿井下掘锚装备智能化是改善行业采掘失衡问题的关键,而掘锚装备的精确定位是实现其智能化的前提。与其他传统定位方法相比,基于视觉的位姿测量方法以其无接触、无累计误差的优势在煤矿井下得到了初步的应用。针对目前煤矿井下掘进工作面掘锚装备视觉定位方法存在的合作标靶结构复杂、标定繁琐的问题,结合掘进工作面原有激光指向仪特征,提出一种基于单激光束信息的掘锚装备视觉定位方法。该方法通过分析激光指向仪光斑及光束图像特征,提出了一种基于二维反正切函数拟合的激光光斑中心提取方法和基于Hough直线检测的激光束中心线提取方法,构建了基于点线特征的双目视觉位姿解算模型,得出了掘锚装备在巷道中的实时位姿。最后,为了验证提出的特征提取方法和视觉定位方法的可行性和准确性,在实验室模拟掘进工作面工况环境搭建平台进行了试验。结果表明:基于矿用激光指向仪信息的掘锚装备视觉定位方法具有较高的位姿测量精度。在50 m的测试范围内,机身位置在巷道坐标系下沿X轴、Y轴和Z轴的平均测量误差分别为25.44、58.64、31.08 mm,其最大误差分别为55.16、127.39、63.57 mm;机身姿态在巷道坐标系下的俯仰角、偏航角和横滚角的平均测量误差分别为0.22°、0.22°、0.41°,其最大误差分别为0.29°、0.37°、0.58°。满足煤矿井下巷道施工的定位精度要求。

基于机器视觉的物流包装条码特征快速提取与识别方法研究

随着技术的进步,特别是机器视觉技术的迅猛发展,条码的读取与识别已经不再依赖于传统的条码扫描器。机器视觉为条码的快速、准确和自动化读取提供了新的技术路径。然而,复杂的物流环境、不同的包装材料以及各种环境因素(如光线、遮挡)会影响机器视觉系统的性能。文章对基于机器视觉的物流包装条码特征快速提取与识别方法展开深入研究,以期为实际物流场景中的条码识别提供一种更为高效和准确的技术手段。

双目视觉的货车车厢完整尺寸测量

针对货车车厢完整尺寸的快速测量,提出了一种基于双目视觉的测量方法。依据相机成像模型和标定原理对双目相机进行标定实验,使用标定后的双目相机从货车侧后方采集货车车厢图像,再采用U 2-Net显著性目标检测算法对采集的货车车厢图像进行车厢轮廓提取,通过改进的基于图像分块的线段检测算法提取车厢边缘,结合车厢轮廓特征利用车厢边缘线段计算出特定的角点坐标,由三维坐标恢复后的车厢角点间距得到车厢尺寸。试验结果表明,车厢尺寸检测的平均误差小于4.06%,体积检测平均误差小于5.83%,可用于实际测量。

一种密集多尺度特征引导代价聚合的改进立体匹配网络

针对目前立体匹配算法在重复纹理、无纹理、边缘等不适定性区域仍存在匹配不准确的问题,提出了一种基于PSMNet的密集多尺度特征引导代价聚合的立体匹配算法—DGNet(Dense multi-scale features Guided aggregation Network)。首先,基于密集连接空洞空间金字塔池化结构设计了密集多尺度特征提取模块,该模块利用不同膨胀率的空洞卷积提取不同尺度的区域级特征,并通过密集连接方式有效整合不同尺度的图像特征,使网络捕获丰富的上下文关系;其次,在每个视差等级下将左右特征图串联形成初始代价体,再提出密集多尺度特征引导代价聚合结构,在聚合代价体的同时自适应融合代价体和密集多尺度特征,从而使后续的解码层在多尺度上下文信息的引导下解码出更加精确和高分辨率的几何信息;最后,将全局优化后的高分辨率代价体送入视差回归模块以获得视差图。实验结果表明:所提算法在KITTI 2015和KITTI 2012数据集上的误匹配率分别降至1.76%和1.24%,SceneFlow数据集上的端点误差降至0.56 px,与GWCNet、CPOP-Net等先进算法相比,所提算法在不适定区域有明显改善。

Efficient Object Segmentation and Recognition Using Multi-Layer Perceptron Networks

Object segmentation and recognition is an imperative area of computer vision andmachine learning that identifies and separates individual objects within an image or video and determines classes or categories based on their features.The proposed system presents a distinctive approach to object segmentation and recognition using Artificial Neural Networks(ANNs).The system takes RGB images as input and uses a k-means clustering-based segmentation technique to fragment the intended parts of the images into different regions and label thembased on their characteristics.Then,two distinct kinds of features are obtained from the segmented images to help identify the objects of interest.An Artificial Neural Network(ANN)is then used to recognize the objects based on their features.Experiments were carried out with three standard datasets,MSRC,MS COCO,and Caltech 101 which are extensively used in object recognition research,to measure the productivity of the suggested approach.The findings from the experiment support the suggested system’s validity,as it achieved class recognition accuracies of 89%,83%,and 90.30% on the MSRC,MS COCO,and Caltech 101 datasets,respectively.

Integrating Transformer and Bidirectional Long Short-Term Memory for Intelligent Breast Cancer Detection from Histopathology Biopsy Images

Breast cancer is a significant threat to the global population,affecting not only women but also a threat to the entire population.With recent advancements in digital pathology,Eosin and hematoxylin images provide enhanced clarity in examiningmicroscopic features of breast tissues based on their staining properties.Early cancer detection facilitates the quickening of the therapeutic process,thereby increasing survival rates.The analysis made by medical professionals,especially pathologists,is time-consuming and challenging,and there arises a need for automated breast cancer detection systems.The upcoming artificial intelligence platforms,especially deep learning models,play an important role in image diagnosis and prediction.Initially,the histopathology biopsy images are taken from standard data sources.Further,the gathered images are given as input to the Multi-Scale Dilated Vision Transformer,where the essential features are acquired.Subsequently,the features are subjected to the Bidirectional Long Short-Term Memory(Bi-LSTM)for classifying the breast cancer disorder.The efficacy of the model is evaluated using divergent metrics.When compared with other methods,the proposed work reveals that it offers impressive results for detection.

基于双目图像深度学习的农作物择优采摘仿真

面对复杂的农作物生长环境,利用传统机器视觉技术采摘易受到未成熟果实以及周围叶片影响,获取的农作物果实图像存在较多的不可用信息,无法完成择优采摘。为了确保采摘的果蔬品质,提出基于双目图像深度学习的农作物择优采摘方法。利用直方图均衡化变换视觉图像区域,明确农作物图像内像素灰度值,均衡化色块不均位置。通过色彩分量调节全局图像颜色,以颜色差异分割双目图像,剔除局部RGB色彩关联性。在提取农作物形状特征前,将深度信息再次归一化,获得作物形态描述符。选择卷积神经网络对图像实行卷积运算,将择优特征结果输入到卷积层内,输出图像分类结果,实现农作物择优采摘。实验结果表明,所提方法的择优采摘精准度达到0.98,果实形状特征识别为0.96。说明所提方法能够准确识别出品质佳的农作物,实现了择优采摘。

改进Census变换与特征融合的立体匹配算法

针对局部立体匹配算法匹配精度较低问题,提出一种改进Census变换与特征融合的立体匹配算法。首先,使用变换窗口的邻域像素信息代替中心像素,解决传统Census变换过度依赖窗口中心像素问题;其次,引入图像的颜色信息与梯度信息构建融合代价计算函数,提高初始匹配代价的可靠性;为建立邻域像素间联系,引入单向动态规划思想进行代价聚合;最后,提出一种基于八方向的视差空洞填充方法对视差图进行优化。实验结果表明:该算法在Middlebury数据集上的平均误匹配率为3.77%,优于其他改进Census变换方法,具有较高的匹配精度。

基于自适应阈值ORB特征提取的果园双目稠密地图构建

针对果园阴暗光照条件下图像特征点匹配数量少、易丢失以及点云稀疏问题,对ORB-SLAM2算法进行了改进,提出了基于自适应阈值ORB特征点提取的果园双目三维地图稠密建图算法。首先在跟踪线程中提出一种自适应阈值的FAST角点提取方法,通过计算不同光照下图像平均像素求解阈值,对左右目图像提取ORB特征,增加了不同光照条件下的特征点匹配数量;然后根据特征点估计相机位姿完成局部地图跟踪,对跟踪线程产生的关键帧地图点进行BA优化完成局部地图构建。在原有算法基础上添加了基于ZED-stereo型相机双目深度融合的稠密建图模块,对左右目关键帧进行特征匹配获得图像对,利用图像对求解深度信息获取地图点,经过深度优化获取相机位姿,根据相机位姿进行局部点云的构建与拼接,最终对获得的点云地图进行全局BA优化,构建果园三维稠密地图。在KITTI数据集序列上进行测试,本文所改进的ORB-SLAM2算法的绝对轨迹误差更加收敛,轨迹误差标准差在00和07序列分别下降60.5%和62.6%,在其他序列上也有不同程度下降,表明本文算法定位精度较原始算法有所提高。不同光照环境下进行算法性能测试,结果表明本文算法较原始算法能更好地适应不同光照条件,在较强光照、正常光照、偏弱光照和阴雨天气下特征点平均匹配数量增加5.32%、4.53%、8.93%、12.91%。进行果园直线和稠密建图试验,结果表明直线行驶偏航角更加收敛,定位精确度高,关键帧提取数量较原始算法下降2.86%、平均跟踪时间减少39.3%;稠密建图效果好,能够很好地反映机器人位姿和果园真实环境信息,满足果园三维稠密点云地图构建需求,可为果园机器人导航路径规划提供支持。