融合脑电-肌电信号的人机协作装配意图识别方法

针对人机协作装配场景下基于单源生理电信号识别协作意图准确率不高、稳定性不好的问题,首先采用支持向量机方法分别从脑电信号和肌电信号识别单源协作装配意图,然后采用D-S证据理论对多源协作装配意图识别结果进行融合,提出了一种融合EEG-EMG信号的人机协作装配意图识别方法。实验结果表明,所提方法可以有效提高人机协作装配意图识别的准确率和稳定性。     

身份证识别系统设计与实现

设计和实现了一个具有数据库管理功能的身份证自动识别系统,能识别身份证图像的文字信息,并实时地与身份证信息数据库进行比对,迅速获得教据库中对应的记录,与文字识别结果进行比对以辨别真伪.实验结果表明,该系统对身份证号码的识别具有很高的准确率,在输入图像质量较好的情况下,对汉字信息的识别率也较高,可以作为身份证信息的录入工具,在旅馆、机场、车站、海关、银行、医院等场合有广泛的应用前景.     

Geometric approach to segmentation and protein localization in cell culture assays

Cell‐based fluorescence imaging assays are heterogeneous and require the collection of a large number of images for detailed quantitative analysis. Complexities arise as a result of variation in spatial nonuniformity, shape, overlapping compartments and scale (size). A new technique and methodology has been developed and tested for delineating subcellular morphology and partitioning overlapping compartments at multiple scales. This system is packaged as an integrated software platform for quantifying images that are obtained through fluorescence microscopy. Proposed methods are model based, leveraging geometric shape properties of subcellular compartments and corresponding protein localization. From the morphological perspective, convexity constraint is imposed to delineate and partition nuclear compartments. From the protein localization perspective, radial symmetry is imposed to localize punctate protein events at submicron resolution. Convexity constraint is imposed against boundary information, which are extracted through a combination of zero‐crossing and gradient operator. If the convexity constraint fails for the boundary then positive curvature maxima are localized along the contour and the entire blob is partitioned into disjointed convex objects representing individual nuclear compartment, by enforcing geometric constraints. Nuclear compartments provide the context for protein localization, which may be diffuse or punctate. Punctate signal are localized through iterative voting and radial symmetries for improved reliability and robustness. The technique has been tested against 196 images that were generated to study centrosome abnormalities. Corresponding computed representations are compared against manual counts for validation.     

Fluorescence histochemical techniques for catecholamines as tools in neurobiology

Formaldehyde-induced and glyoxylic-acid-induced fluorescence histochemistry permits the tissue localization of catecholamines in the central nervous system (CNS) and peripheral nervous system (PNS), and in culture. Counterstains such as ethidium bromide provide excellent background identification of specific innervated regions in both the CNS and the periphery. Use of fluorescence histochemistry with immunocytochemistry can elucidate catecholamine-peptide relationships. Gelatin-ink perfusion used with fluorescence histochemistry permits the investigation of neuro-vascular relationships and documentation of vascular and parenchymal compartmentation of innervation. Combined use of fluorescence histochemistry and retrograde tracing methods demonstrates the specific cellular sources of innervation of target regions. Micropunch neurochemical analysis provides quantitative data for correlation with fluorescence histochemistry within a target region of innervation, and micro-spectrofluorometric analysis provides a semi-quantitative evaluation of the amount of fluorophore within a target region or within specific subcellular compartments such as the cell body or terminals.     

基于机器学习识别与标记分水岭分割的盲道图像定位

视觉导盲仪是一种旨在解决盲人出行困难的导盲设备,为了实现视觉导盲仪诱导盲人找到盲道并沿盲道行走,提出了一种基于机器学习识别与标记分水岭分割的盲道图像定位算法,通过离线训练与在线的识别、分割来定位盲道区域。首先对盲道图像进行视角变换的预处理,根据盲道的地面方程将变化的倾斜视角转换为固定的俯视视角,消除射影变换带来的失真;然后利用局部二进制模式描述子提取鸟瞰图的纹理特征,以自适应增强学习算法离线训练盲道识别分类器;进而利用分类器对鸟瞰图像进行在线识别,粗略确定盲道区域;最后将识别结果进行形态学处理后作为标记,利用标记分水岭算法得到精确分割的盲道区域并定位盲道中心线。在研制的视觉导盲仪上进行验证,结果表明盲道定位准确率达到了95.44%,速度平均每秒8帧,具有高准确率的同时达到实时性要求,为盲道的准确三维定位提供了必要条件。     

基于稀疏编码空间金字塔匹配和GA-SVM的列车故障自动识别

针对货车运行故障动态图像中车辆挡键、集尘器和安全链锁紧螺栓的故障检测,提出一种基于稀疏编码空间金字塔匹配和遗传算法优化的支持向量机相结合的通用故障自动识别算法。首先在不同尺度空间对样本图像进行划分,对每个部分提取尺度不变特征变换特征,利用随机抽取样本的SIFT特征通过迭代学习生成字典并进行稀疏编码;其次利用主成分分析定义编码后的特征对故障识别准确率的贡献值,并据此对编码特征进行降维;然后利用编码降维后的特征结合遗传算法对线性SVM分类器进行训练;最后用训练好的分类器模型对挡键、集尘器和安全链锁紧螺栓的故障进行识别。实验结果表明,本文提出的算法能较好的应用于3种不同类型的故障识别,识别率分别为97.25%、99.00%和97.50%,同时对噪声和光照变化具有一定的鲁棒性,能满足车辆故障的实际检测需求。     

输电线巡检机器人障碍物定位与识别

针对巡检机器人运行中对线路障碍物的自主定位与识别,分析了巡检机器人自主运行中对障碍物信息感知的精度和实时性要求,提出了一种用于线路障碍物定位与识别的多传感器集成结构,并介绍了各个阶段的障碍物定位与识别算法。实验表明该方法可以实现机器人稳定自主行驶、可靠定位与识别障碍物,满足各阶段机器人对障碍物信息感知的精度和实时性要求,从而为机器人制定自主越障策略提供信息。     

基于yolov3的车牌定位识别系统

为解决目前市场上车牌识别设备的高时延以及对高性能处理器高依赖性的问题,提出了一种基于yolov3的车牌定位识别方法,实现对车辆的车牌进行自动定位与识别。该方法通过收集5000张车辆图像数据,打包为VOC数据集,使用LabelImg工具对图像进行标注,并构建13个卷积层的yolo模型,在loss曲线趋于稳定之后,完成模型的训练,最终使用已训练的模型对车辆图像数据进行车牌识别。实验表明:在1000个测试数据中,该方法识别率达97.5%,平均耗时18.19 ms,能够快速精准的对车辆进行车牌识别。     

基于BP神经网络的电缆沟监控模式识别方法

为了减少电缆沟盲目施工与偷盗等因素造成的损失和影响，引入一种智能的线路监控技术。该技术基于分布式光纤传感技术，利用BP神经网络对光纤传感器的四种常见的扰动信号进行模式识别。通过将光纤传感器得到的时域扰动信号用特定的程序处理后转换成图像，再经特定的图像处理流程后形成模式识别的样本。利用这些样本训练BP神经网络，并将训练好的模型应用到实际的电缆沟安全监控系统中进行测试。测试结果表明，电缆沟的总体识别成功率为98.16%，该识别方法还可以应用于光纤周界安防系统等领域。     

基于Faster R-CNN的工件螺纹孔目标检测

为了提高螺纹孔目标检测的准确率,结合双相机视觉系统与Hough变换圆检测算法,提出了一种基于Faster R-CNN的螺纹孔目标检测方法。首先建立了由双相机组成的图像获取系统,通过安置在高处的工业相机采集工件整体图像,利用Hough变换圆检测算法初步筛选出工件上的疑似螺纹孔的位置,并驱动第二个工业相机逐个在近处采集经Hough变换检测出的疑似螺纹孔的局部精确图像。然后,在自建的螺纹孔数据集上训练以ResNet50为基础网络的Faster R-CNN目标检测模型。最后,将螺纹孔处局部图像输入训练好的Faster R-CNN目标检测模型进一步识别并进行定位。实验结果表明,该方法能有效地避免螺纹孔小目标检测,相对于单独使用Hough变换方法或者Faster R-CNN目标检测方法检测螺纹孔,具有更高的识别和定位精度。     

Quantitative measurement of mRNA at different loci within an individual living cell

Asymmetric localizations of cellular proteins and mRNAs are important for cell functions such as division, differentiation and development. The localization of specific mRNA generates cell polarity by controlling the translation sites of specific proteins and thereby restricting their locations to appropriate cellular regions. We have previously reported a novel method based on atomic force microscopy (AFM) for examining gene expression in a single living cell without killing or destroying it. An AFM tip was inserted into a living cell to extract mRNAs, which were analyzed after multiplication by RT-PCR and quantitative PCR. By applying this method, in this study we performed quantitative measurement of mRNA at different loci within individual living cells.     

Combining ocFLIM and FIDSAM reveals fast and dynamic physiological responses at subcellular resolution in living plant cells

For a deeper understanding of molecular mechanisms within cells and for the realization of predictive biology for intracellular processes at subcellular level, quantitative biology is required. Therefore, novel optical and spectroscopic technologies with quantitative and dynamic output are needed in cell biology. Here, we present a combined approach of novel one-chromophore fluorescence lifetime imaging microscopy to probe the local environment of fluorescent fusion proteins and fluorescence intensity decay shape analysis microscopy to suppress interfering autofluorescence. By applying these techniques, we are able to analyse the subcellular localization and partitioning of a green fluorescence protein fusion of the salt stress-induced protein low temperature induced (LTI)6b in great detail with high spatial and temporal resolution in living cells of Arabidopsis plants.     

Multi-modal Gesture Recognition using Integrated Model of Motion,Audio and Video

Gesture recognition is used in many practical applications such as human-robot interaction, medical rehabilitation and sign language. With increasing motion sensor development, multiple data sources ha...     

质谱多反应监测技术（MRM）及其在蛋白质细胞器定位与定量研究中的应用

肝脏中的蛋白质只有处于特定的亚细胞中才能发挥其功能。研究特定蛋白质的亚细胞定位对于肝脏功能研究有重要意义[1]。对于多定位的蛋白质,确定其在不同细胞器中量的变化,也是肝脏蛋白表达谱研究的重要内容之一。目前,蛋白亚细胞定位分析方法主要包括实验方法和理论预测方法两大类。其中实验方法主要有绿色荧光蛋白（GFP）-激光共聚焦显微镜法、免疫电镜胶体金标记法、差速离心-免疫印迹法等。理论预测方法主要根据蛋白质的氨基酸组成、氨基酸对组成以及位置特异性打分矩阵等分类特征,采用模糊k近邻、支持向量机及分组重量编码法等方法进行理论预测。实验方法不仅分析通量低,还需要特异的抗体,并且在定位的同时难以实现定量分析。对于不同的理论预测方法而言,往往会得出不同的结论,准确度差。因而需要建立一种可以实现通量化、准确地进行蛋白亚细胞定位及定量分析的方法。 近年来,质谱多反应监测技术（MRM）在蛋白质定量分析方面的应用越来越广泛^[2]。该技术通过特异性的检测预先设置的母离子和子离子,在母离子和子离子两个水平排除其它离子的检测干扰, 从而增强检测的特异性。同时结合同位素稀释技术可以同时实现对多种目标蛋白质的绝对定量。由于该技术采用特异性检测预设的母子离子对的方法,从而保证了检测的重现性。而依据特定子离子和相应的同位素标记子离子对峰面积进行绝对定量的方法保证了定量的可靠性。同时可以批量化同时进行多个蛋白质的亚细胞定量及定位分析。因而本研究利用MRM技术,用于目标蛋白在肝脏细胞器中定位及定量分析研究。 鉴于本研究样品采用蔗糖密度梯度离心法制备,首先对细胞器样品的分离效果进行评价。选用文献报道^[3]的正向和反向免疫评价抗体蛋白,包括线粒体Marker蛋白COX4、VDAC和Cytc,细胞核Marker蛋白Lamin B,质膜Marker蛋白Na＋K＋-ATPase等。利用MRM技术分析蛋白在不同细胞器样品中的含量,从而对细胞器样品纯度进行评价。同时考察方法的灵敏度,重现性以及动态范围。 在确定了细胞器样品纯度的基础上,利用MRM质谱技术对选定的部分目标蛋白进行了定位分析。根据这些蛋白在不同细胞器样品中的含量,判断其在不同细胞器样品中的定位。并选择其中一些蛋白进行GFP-激光共聚焦显微镜分析,与基于MRM技术的定位分析结果进行对比验证。     

A multi-sensor relation model for recognizing and localizing faults of machines based on network analysis

Recently, advanced sensing techniques ensure a large number of multivariate sensing data for intelligent fault diagnosis of machines. Given the advantage of obtaining accurate diagnosis results, multi-sensor fusion has long been studied in the fault diagnosis field. However, existing studies suffer from two weaknesses. First, the relations of multiple sensors are either neglected or calculated only to improve the diagnostic accuracy of fault types. Second, the localization for multi-source faults is seldom investigated, although locating the anomaly variable over multivariate sensing data for certain types of faults is desirable. This article attempts to overcome the above weaknesses by proposing a global method to recognize fault types and localize fault sources with the help of multi-sensor relations (MSRs). First, an MSR model is developed to learn MSRs automatically and further obtain fault recognition results. Second, centrality measures are employed to analyze the MSR graphs learned by the MSR model, and fault sources are therefore determined. The proposed method is demonstrated by experiments on an induction motor and a centrifugal pump. Results show the proposed method’s validity in diagnosing fault types and sources.     

基于机器视觉的芯片编号识别算法

针对芯片检测过程中芯片编号形状小、人眼无法识别的现状,开发了一套芯片编号识别系统。系统以机器视觉技术为基础,配备高分辨率Basler相机、高倍率镜头以及高性能点光源,能实现高分辨率图像采集;利用Hough直线检测对图像进行水平旋转,利用图像轮廓查找算法对兴趣区域进行提取分割,利用帧差法对兴趣区域与标准库进行比较,找到最匹配的图像,其标签为识别结果。实验结果表明,运用系统的图像处理算法可以代替人眼通过高倍放大镜识别,改进了芯片检测工艺,大大提高了芯片检测的效率。     

Three-dimensional object recognition system for enhancing the intelligence of a KUKA robot

Machine intelligence has been a research hotspot in mechatronics in recent years. This research presents a 2D/3D object recognition system for enhancing the intelligence of an industrial robot (KUKA robot). The image processing and object recognition algorithms were developed using software packages VisionPro and LabVIEW. Experiments were carried out to verify the performance of the system. It can be concluded that the system is able to recognise any general 2D object within a time of six seconds. The performance of the system in 3D object recognition is slower compared to 2D objects, which is largely affected by the number of trained images stored in the database, the complexity of the object, and the presence of similar objects in the database. Despite the complexity of the objects being recognised, both the overall accuracy and success rate of the system are close to 100%. The developed system proved to be robust and allows for automatic recognition of objects in the manufacturing environment described in this paper.     

一种基于加速计的手势识别算法及其在虚拟环境中的应用

针对虚拟环境下的三维交互问题,提出了基于加速计的手势识别算法。识别之前先对加速计输出做量化处理,再用离散型隐马尔科夫模型(HMM)进行建模、训练与识别。基于此识别算法,将手势作为一种交互方式应用在三维交互系统中,建立了手势-语义映射表,以训练标准手势,最后对输入手势进行了识别,根据识别结果完成了交互语义。将算法应用于虚拟奥运博物馆实例中,实验结果表明,该算法可有效快速识别出用户输入手势,且平均识别率达到90%以上,用户凭借此方法能准确无误地完成交互。     

一种基于深度学习图像超分的环形靶标稳定检测方法

为提高远距离、大倾角条件下环形靶标的识别率与定位精度,提出了一种基于深度学习图像超分的环形靶标稳定检测方法。通过真实图像与合成图像的混合数据集来构建多角度、多距离的图像空间集合,采用像素损失与感知损失来改进图像超分辨率模型的损失函数,从而实现图像的高分辨率重建,丰富靶标轮廓的图像细节,利用已训练好的图像超分模型重建图像,最后使用传统的检测算法识别与定位环形靶标。实验结果表明,环形靶标识别率可提高40%,靶标定位精度可提高8.47%。     

基于弹性网正则化的广义逆波束形成

广义逆波束形成是一种高效的声源识别方法。然而受限于较低的算法稳健性,使得其难以实现高精度的声源识别定位。为提高广义逆波束形成声源识别性能,结合弹性网正则化方法和广义逆理论提出一种基于弹性网正则化的广义逆波束形成。首先从广义逆理论出发介绍了特征向量求解以及阈值截断滤波过程,并结合弹性网正则化思想全面阐述了基于弹性网正则化的广义逆波束形成基本理论;其次建立了数值仿真模型,以单极子和多声源识为研究对象,对比其他波束形成算法详细分析了声源类型与频率等因素对其声源成像性能的影响。最后以单极子、不相干以及相干声源为研究对象进行实验分析,结果表明由于基于弹性网正则化的广义逆波束形成的波束输出解具有较强的稀疏性和稳健性,使得其相比传统广义逆波束形成,能更精准地识别定位声源。