An FPGA Design for Real-Time Image Denoising

The increasing use of images in miscellaneous applications such as medical image analysis and visual quality inspection has led to growing interest in image processing. However, images are often contaminated with noise which may corrupt any of the following image processing steps. Therefore, noise filtering is often a necessary preprocessing step for the most image processing applications. Thus, in this paper an optimized field-programmable gate array (FPGA) design is proposed to implement the adaptive vector directional distance filter (AVDDF) in hardware/software (HW/SW) codesign context for removing noise from the images in real-time. For that, the high-level synthesis (HLS) flow is used through the Xilinx Vivado HLS tool to reduce the design complexity of the HW part. The SW part is developed based on C/C++ programming language and executed on an advanced reduced instruction set computer (RISC) machines (ARM) Cortex-A53 processor. The communication between the SW and HW parts is achieved using the advanced extensible Interface stream (AXI-stream) interface to increase the data bandwidth. The experiment results on the Xilinx ZCU102 FPGA board show an improvement in processing time of the AVDDF filter by 98% for the HW/SW implementation relative to the SW implementation. This result is given for the same quality of image between the HW/SW and SW implementations in terms of the normalized color difference (NCD) and the peak signal to noise ratio (PSNR).     

PredMaX: Predictive maintenance with explainable deep convolutional autoencoders

A novel data exploration framework (PredMaX) for predictive maintenance is introduced in the present paper. PredMaX offers automatic time period clustering and efficient identification of sensitive machine parts by exploiting hidden knowledge in high-dimensional, unlabeled temporal data. Condition monitoring systems often provide such data, which is further analyzed by human experts or used for training predictive models.PredMaX reduces data dimensionality in two steps: An explainable deep convolutional autoencoder is applied on the data first, followed by principal component analysis. The automatic clustering is performed in the latent space of the autoencoder, ensuring higher accuracy than the clustering in the space of principal components. If clusters of normal and abnormal operation are known, the reasoning module is able to reveal the measurement channels that contributed the most to the latent representation moving from normal to abnormal operation.Beyond the detailed presentation of the PredMaX approach, the paper presents the case study of identifying the most important signals that can be used for predicting oil degradation in an industrial gearbox. The case study is performed on a data-driven basis with minimal human assistance and without preliminary knowledge of the machine.     

Image denoising in extended Shearlet domain using hidden Markov tree models

Denoising of images is one of the most basic tasks of image processing. It is a challenging work to design an edge-preserving image denoising scheme. Extended discrete Shearlet transform (extended DST) is an effective multi-scale and multi-direction analysis method; it not only can exactly compute the Shearlet coefficients based on a multiresolution analysis, but also can provide nearly optimal approximation for a piecewise smooth function. In this paper, a new image denoising approach in extended Shearlet domain using hidden Markov tree (HMT) model is proposed. Firstly, the joint statistics and mutual information of the extended DST coefficients are studied. Then, the extended DST coefficients are modeled using an HMT model with Gaussian mixtures, which can effectively capture the intra-scale and inter-scale dependencies. Finally, the extended Shearlet HMT model is applied to image denoising. Extensive experimental results demonstrate that our extended Shearlet HMT denoising method can obtain better performances in terms of both subjective and objective evaluations than other state-of-the-art HMT denoising techniques. Especially, the proposed method can preserve edges very well while removing noise.     

基于希尔伯特黄变换的刀具磨损特征提取

概述了希尔伯特黄变换(HHT)的基本理论和算法,对信号经过经验模态分解(EMD)后得到的固有模态函数(IMF)求取振幅均值,差值筛选出与刀具磨损相关的IMF分量,并对单分量固有模态函数求取边际谱,获取边际谱最大幅值点,建立他们与刀具磨损之间的映射关系,进行特征提取,将其作为神经网络的输入特征向量,结合希尔伯特三维时频谱进行刀具磨损状态的判断。研究结果证明,该方法可以作为刀具磨损监测中信号特征提取的一种简单和可靠的方法。     

增量深度学习目标跟踪

由于现有目标跟踪算法在复杂环境下易发生目标漂移甚至跟踪丢失,故本文提出了以双重采样粒子滤波为框架,基于增量深度学习的目标跟踪算法。该算法在粒子滤波中引入粒子集规模自适应调整的双重采样来解决粒子衰减及贫化问题,并利用无监督特征学习预训练深度去噪自编码器以克服跟踪中训练样本的不足。将深度去噪自编码器应用到在线跟踪中,使提取的特征集合能够有效表达粒子图像区域。在深度去噪自编码器中添加了增量特征学习方法,得到了更有效的特征集以适应跟踪过程中目标外观变化。该方法还用线性支持向量机对特征集合进行分类,提高对粒子集合的分类精度,以得到更精确的目标位置。在复杂环境下对不同图片序列进行的实验表明:该算法的跟踪综合评价指标为94%、重叠率为74%,平均帧率为13frame/s。与现有的跟踪算法相比,本算法有效地解决目标漂移甚至跟踪丢失问题,并且对遮挡、相似背景、光照变化、外观变化具有更好的鲁棒性及精确度。     

基于NSCT扩散的图像去噪增强方法

本文提出了一种基于非下采样Contourlet变换与非线性各向异性扩散的方法进行含噪图像的去噪和增强。首先对含噪图像进行非下采样Contourlet分解,对每个分解层的各个子带进行非线性收缩和拉伸,以达到抑制噪声和增强图像特征的目的。然后,对去噪增强后图像的Contourlet小系数进行空间域的非线性各向异性扩散,以去除由于进行非下采样Contourlet去噪所造成的为伪Gibbs现象和 side-band效应。实验结果表明,本文方法相比于无扩散的Wavelet和Contourlet方法相比,不仅对图像进行了去噪和增强,而且有效的抑制了伪Gibbs现象和 side-band效应。     

基于深度神经网络的多电平逆变器故障诊断

为解决多电平逆变器的故障诊断问题,文中提出了一种基于深度神经网络的多电平级联H桥逆变器的故障诊断方法。首先,介绍了多电平级联H桥逆变器的故障模型;然后,采用基于堆栈自编码器的深度神经网络直接从故障原始数据中进行故障特征提取;最后,运用SOFTMAX分类器对故障特征数据进行分类从而实现多电平逆变器的故障检测与诊断。基于MATLAB/Simulink对本文提出的方法进行了仿真实验,该方法与传统的智能诊断算法相比具有更高的准确率和更好的鲁棒性。     

基于深度自编码的大曲高光谱数据解混研究

利用高光谱技术检测大曲发酵品质时,获取的水分等高含量物质的高光谱数据可能掩盖对大曲质量评价至关重要的微量物质高光谱数据。为方便后续更微量物质的光谱曲线分解,需先排除水分等高含量物质的数据干扰,该文通过建立无监督的深度自编码模型,可实现大曲水分高光谱曲线分解。通过实验设计,采集与水混合后的成品曲粉光谱数据。首先编码部分,将混合大曲光谱曲线压缩为低维表示,即端元;解码部分,将光谱的低维表示解压重构为原始光谱曲线,结合比较不同目标函数,反向传递重构误差,更新解码权重;最终,通过端元解码出水与曲粉各自的光谱曲线,运用欧氏距离与皮尔逊相关系数方法从特征距离和相关系数两方面同时评价解混效果。实验显示,利用深度自编码解混模型,选择L-C目标函数得到的解混效果最好,解混曲线与纯曲粉曲线的欧式距离与皮尔逊相似度分别为0.3427和0.9967。研究表明,利用深度自编码网络能够对大曲高光谱数据进行解混,可为实现大曲高光谱微量物质检测提供理论支持和技术支撑。     

Recent Advances in Data Mining Techniques and Their Applications in Hyperspectral Image Processing for the Food Industry

Hyperspectral imaging (HSI) facilitates better characterization of intrinsic and extrinsic properties of foods by integrating traditional spectral and image techniques, in which careful and sophisticated data processing plays an important role. In the past decade, much progress has been made on applying various algorithms to deal with hyperspectral images. This review first introduces the general procedure of hyperspectral data analysis and then illustrates the most typically and commonly used algorithms for denoising, feature selection, model establishment, and evaluation, as well as their applications for assessing food quality, safety, and authenticity. Finally, brief summaries for regression and classification methods are presented. This article will provide a guideline for data mining in the future development of HSI in the food field.     

EMD降噪和关联维数相结合在滚动轴承故障诊断中的应用

针对滚动轴承产生的非线性振动信号的特点,提出将EMD降噪和关联维数相结合来描述轴承工作状态进而对其进行故障诊断的方法.通过实验定量计算出EMD降噪前后滚动轴承正常工况和不同部件故障时的关联维数值,结果表明,经EMD降噪处理后,不同模式下的关联维数具有明显的可分性和严格的区间范围,可以作为智能监测与诊断的参考依据.