Early-stage atherosclerosis detection using deep learning over carotid ultrasound images

This paper proposes a computer-aided diagnosis tool for the early detection of atherosclerosis. This pathology is responsible for major cardiovascular diseases, which are the main cause of death worldwide. Among preventive measures, the intima-media thickness (IMT) of the common carotid artery stands out as early indicator of atherosclerosis and cardiovascular risk. In particular, IMT is evaluated by means of ultrasound scans. Usually, during the radiological examination, the specialist detects the optimal measurement area, identifies the layers of the arterial wall and manually marks pairs of points on the image to estimate the thickness of the artery. Therefore, this manual procedure entails subjectivity and variability in the IMT evaluation. Instead, this article suggests a fully automatic segmentation technique for ultrasound images of the common carotid artery. The proposed methodology is based on machine learning and artificial neural networks for the recognition of IMT intensity patterns in the images. For this purpose, a deep learning strategy has been developed to obtain abstract and efficient data representations by means of auto-encoders with multiple hidden layers. In particular, the considered deep architecture has been designed under the concept of extreme learning machine (ELM). The correct identification of the arterial layers is achieved in a totally user-independent and repeatable manner, which not only improves the IMT measurement in daily clinical practice but also facilitates the clinical research. A database consisting of 67 ultrasound images has been used in the validation of the suggested system, in which the resulting automatic contours for each image have been compared with the average of four manual segmentations performed by two different observers (ground-truth). Specifically, the IMT measured by the proposed algorithm is 0.625 ± 0.167 mm (mean ± standard deviation), whereas the corresponding ground-truth value is 0.619 ± 0.176 mm. Thus, our method shows a difference between automatic and manual measures of only 5.79 ± 34.42 μm. Furthermore, different quantitative evaluations reported in this paper indicate that this procedure outperforms other methods presented in the literature.     

小样本条件下的RGB-D显著性物体检测

目的 现有基于RGB-D(RGB-depth)的显著性物体检测方法往往通过全监督方式在一个较小的RGB-D训练集上进行训练,因此其泛化性能受到较大的局限。受小样本学习方法的启发,本文将RGB-D显著性物体检测视为小样本问题,利用模型解空间优化和训练样本扩充两类小样本学习方法,探究并解决小样本条件下的RGB-D显著性物体检测。方法 模型解空间优化通过对RGB和RGB-D显著性物体检测这两种任务进行多任务学习,并采用模型参数共享的方式约束模型的解空间,从而将额外的RGB显著性物体检测任务学习到的知识迁移至RGB-D显著性物体检测任务中。另外,训练样本扩充通过深度估计算法从额外的RGB数据生成相应的深度图,并将RGB图像和所生成的深度图用于RGB-D显著性物体检测任务的训练。结果 在9个数据集上的对比实验表明,引入小样本学习方法能有效提升RGB-D显著性物体检测的性能。此外,对不同小样本学习方法在不同的RGB-D显著性物体检测模型下(包括典型的中期融合模型和后期融合模型)进行了对比研究,并进行相关分析与讨论。结论 本文尝试将小样本学习方法用于RGB-D显著性物体检测,探究并利用两种不同小样本...     

Flexible robot sealant dispensing cell using RGB-D sensor and off-line programming

In aerospace manufacture the accurate and robust application of sealant is an integral and challenging part of the manufacturing process that is still performed by human operator. Automation of this process is difficult and not cost effective due to the high variability in the parts to operate and also the difficulty associated with programming industrial robotic systems. This work tries to overcome these problems by presenting an AOLP (Automatic Off-Line Programming) system for sealant dispensing through the integration of the ABB's proprietary OLP (Off-Line Programming) system RobotStudio with a relatively new RBG-D sensor technology based on structured light and the development of a RobotStudio add-on. The integration of the vision system in the generation of the robot program overcomes the current problems related to AOLP systems that rely on a known model of the work environment. This enables the ability to dynamically adapt the model according to sensor data, thus coping with environmental and parts variability during operation. Furthermore it exploits the advantages of an OLP system simplifying the robot programming allowing for faster automation of the process.     

A water-immersible 2-axis scanning mirror microsystem for ultrasound andha photoacoustic microscopic imaging applications

Fast scanning is highly desired for both ultrasound and photoacoustic microscopic imaging, whereas the liquid environment required for acoustic propagation limits the usage of traditional microelectromechanical systems (MEMS) scanning mirrors. Here, a new water-immersible scanning mirror microsystem has been designed, fabricated and tested. To achieve reliable underwater scanning, flexible polymer torsion hinges fabricated by laser micromachining were used to support the reflective silicon mirror plate. Two efficient electromagnetic microactuators consisting of compact RF choke inductors and high-strength neodymium magnet disc were constructed to drive the silicon mirror plate around a fast axis and a slow axis. The performance of this water-immersible scanning mirror microsystem in both air and water were tested using the laser tracing method. For the fast axis, the resonance frequency reached 224 Hz in air and 164 Hz in water, respectively. The scanning angles in both air and water under ±16 V DC driving were ±12°. The scanning angles in air and water under ±10 V AC driving (at the resonance frequencies) were ±13.6° and ±10°. For the slow axis, the resonance frequency reached 55 Hz in air and 38 Hz in water, respectively. The scanning angles in both air and water under ±10 V DC driving were ±6.5°. The scanning angles in air and water under ±10 V AC driving (at the resonance frequencies) were ±8.5° and ±6°. The feasibility of using such a water-immersible scanning mirror microsystem for scanning ultrasound microscopic imaging has been demonstrated with a 25-MHz ultrasound pulse/echo system and a target consisting of three optical fibers.     

Implementation of phase-locked loop control for MEMS scanning mirror using DSP

The present study has employed the phase-locked loop control method to ensure the operating of MEMS actuators at their resonant frequency. In this study, the control algorism was simulated by the MATLAB. Further, the digital signal processing (DSP) technique was adopted to implement the concept of phase-locked loop control algorithm. Thus, a wide VCO lock-in dynamic range was achieved. In applications, the optical scanners fabricated using the SOI wafer and MOSBE process were respectively employed to demonstrate the present technique. The test show that the resonant frequency was tracked for various driving voltages, loop gains, and initial frequency offsets. Thus, the variation of the scanning angle resulted from the offset of the resonant frequency of the devices can be prevented.     

面向RGB-D场景解析的三维空间结构化编码深度网络

有效的RGB-D图像特征提取和准确的3D空间结构化学习是提升RGB-D场景解析结果的关键。目前,全卷积神经网络（FCNN）具有强大的特征提取能力,但是,该网络无法充分地学习3D空间结构化信息。为此,提出了一种新颖的三维空间结构化编码深度网络,内嵌的结构化学习层有机地结合了图模型网络和空间结构化编码算法。该算法能够比较准确地学习和描述物体所处3D空间的物体分布。通过该深度网络,不仅能够提取包含多层形状和深度信息的分层视觉特征（HVF）和分层深度特征（HDF）,而且可以生成包含3D结构化信息的空间关系特征,进而得到融合上述3类特征的混合特征,从而能够更准确地表达RGB-D图像的语义信息。实验结果表明,在NYUDv2和SUNRGBD标准RGB-D数据集上,该深度网络较现有先进的场景解析方法能够显著提升RGB-D场景解析的结果。     

Color smoothing for RGB-D data using entropy information

RGB-D sensors are capable of providing 3D points (depth) together with color information associated with each point. These sensors suffer from different sources of noise. With some kinds of RGB-D sensors, it is possible to pre-process the color image before assigning the color information to the 3D data. However, with other kinds of sensors that is not possible: RGB-D data must be processed directly. In this paper, we compare different approaches for noise and artifacts reduction: Gaussian, mean and bilateral filter. These methods are time consuming when managing 3D data, which can be a problem with several real time applications. We propose new methods to accelerate the whole process and improve the quality of the color information using entropy information. Entropy provides a framework for speeding up the involved methods allowing certain data not to be processed if the entropy value of that data is over or under a given threshold. The experimental results provide a way to balance the quality and the acceleration of these methods. The current results show that our methods improve both the image quality and processing time, as compared to the original methods.     

Robust sensor faults detection for induction motor using observer

Current sensor is one of the key elements in the control system of induction motor.Whether the accurate measurement of variables reflecting motor operation status can be made will directly affect the control effect on motor system and therefore the timely,accurate detection of sensor fault is necessary.This paper brings forward an observerbased method of residual generation and fault detection on the basis of the mathematical model of the induction motor.As whether or not the nonlinear part satisfies the Lipschitz conditions does not limit the observer design,the application of such an observer is expanded.Meanwhile,the contradiction between robust error and fault sensitivity is also settled.The correctness and effectiveness of such method are verified by experimental testing on the simulated fault which also casts light on engineering practice.     

Automated segmentation of RGB-D images into a comprehensive set of building components using deep learning

Building information modeling (BIM) has a semantic scope that encompasses all building systems, e.g. architectural, structural, mechanical, electrical, and plumbing. Automated, comprehensive digital modeling of buildings will require methods for semantic segmentation of images and 3D reconstructions capable of recognizing all building component classes. However, prior building component recognition methods have had limited semantic coverage and are not easily combined or scaled. Here we show that a deep neural network can semantically segment RGB-D (i.e. color and depth) images into 13 building component classes simultaneously despite the use of a small training dataset with only 1490 object instances. For this task, the method achieves an average intersection over union (IoU) of 0.5. The dataset was designed using a common building taxonomy to ensure comprehensive semantic coverage and was collected from a diversity of buildings to ensure intra-class diversity. As a consequence of its semantic scope, it was necessary to perform pre-segmentation and 3D to 2D projection as leverage for dataset annotation. In creating our deep learning pipeline, we found that transfer learning, class balancing, and prevention of overfitting effectively overcame the dataset’s borderline adequate class representation. Our results demonstrate how the semantic coverage of a building component recognition method can be scaled to include a larger diversity of building systems. We anticipate our method to be a starting point for broadening the scope of the semantic segmentation methods involved in digital modeling of buildings.     

MEMS扫描镜光学性能的圆片级自动检测系统设计

为快速准确地检测出一张硅片上数百个MEMS扫描镜的良品与次品,设计了一种自动测量系统.选用高精度二维位置敏感探测器(PSD)设备,设计其后置硬件电路并编写数据采集的程序.结合经典三角法测量角度的原理搭建自动检测系统,在室温和正常光照条件下,完成了MEMS扫描镜光学性能圆片级自动检测的实验.结果表明:该系统能够快速准确地同时对扫描镜的两个轴进行检测并给出结果.     

Real-time obstacle detection for legged robots using the Kinect sensor

Autonomous navigation of legged robots in complex environments poses a great deal of challenges compared with ground vehicles because of their different terrain traverse capabilities. An obstacle for vehicles may be traversable for legged robots. This paper proposes a real-time obstacle detection algorithm for legged robots using the Microsoft Kinect sensor. First, the elevation map of a reference grid is calculated. Then an obstacle definition for legged robots is proposed, which makes it possible for a legged robot to discriminate traversable areas from non-traversable areas. To reduce computational cost, sometimes, efficient judging rules are developed to identify obstacles. A spiral search strategy is proposed to find the most ground-like point as the starting point for graph-based traversal. Breadth-First-Traversal of the graph is used to label all traversable areas connecting to the starting point. Experimental results demonstrate that our algorithm is reliable and efficient. The proposed algorithm can be employed in real-time obstacle detection for legged robots in complex environments.     

基于深度置信网络的Android恶意软件检测

针对采用重打包和代码混淆技术的Android恶意软件检测准确率低的问题,提出了一种基于深度置信网络的Android恶意软件检测算法。通过自动化提取Android应用软件的特征,构建对应的特征向量,训练基于深度置信网络的深度学习模型,实现了一种新的基于深度置信网络的Android恶意软件检测算法。实验结果表明,基于深度置信网络的深度学习模型可以更好地表征Android恶意软件,其检测效果也明显优于传统的机器学习模型。     

基于深度学习的安卓恶意应用检测

针对传统安卓恶意程序检测技术检测准确率低,对采用了重打包和代码混淆等技术的安卓恶意程序无法成功识别等问题,设计并实现了DeepDroid算法。首先,提取安卓应用程序的静态特征和动态特征,结合静态特征和动态特征生成应用程序的特征向量;然后,使用深度学习算法中的深度置信网络（DBN）对收集到的训练集进行训练,生成深度学习网络;最后,利用生成的深度学习网络对待测安卓应用程序进行检测。实验结果表明,在使用相同测试集的情况下,DeepDroid算法的正确率比支持向量机（SVM）算法高出3.96个百分点,比朴素贝叶斯（Naive Bayes）算法高出12.16个百分点,比K最邻近（KNN）算法高出13.62个百分点。DeepDroid算法结合了安卓应用程序的静态特征和动态特征,采用了动态检测和静态检测相结合的检测方法,弥补了静态检测代码覆盖率不足和动态检测误报率高的缺点,在特征识别的部分采用DBN算法使得网络训练速度得到保证的同时还有很高的检测正确率。     

一种电容传感器金属材料表面缺陷检测方法

基于电容传感器原理,实现了一种简单有效的金属材料表面缺陷检测方法.首先介绍了电容传感器的工作原理,描述了单电极传感器用于检测金属材料表面缺陷时的基本方法,并给出了相应的电路模型.设计了系列实验验证该检测方法的可行性.实验结果表明：电容传感器对金属材料表面缺陷较敏感,通过单片集成电场成像集成器件MC33794能够快速简单实现该检测方法,为金属材料表面缺陷（如腐蚀）提供了一种快速、便捷、有效的检测方法.     

WGI-Net: A weighted group integration network for RGB-D salient object detection

Computational Visual Media - Salient object detection is used as a pre-process in many computer vision tasks (such as salient object segmentation, video salient object detection, etc.). When...     

Pig weight prediction system using RGB-D sensor and AR glasses: analysis method with free camera capture direction

Artificial Life and Robotics - Pig weights are important indicator for the healthcare and the economic operation of pig farms, and the development of a system to easily estimate these weights is...     

Decompression and speckle detection for ultrasound images using the homodyned k-distribution

The ultrasound envelope intensity distribution can be used for speckle detection and for measuring the distance between images by speckle decorrelation. However, this intensity signal is rarely available. Many researchers work with B-scan data which has been scan-converted and subject to nonlinear mappings to compress the dynamic range. This paper presents an approximate algorithm for recovering the intensity signal from B-scan data. It is then used as the basis of a speckle detector using the statistics of a homodyned k-distribution.     

Image steganography using deep learning based edge detection

This paper introduces a deep learning-based Steganography method for hiding secret information within the cover image. For this, we use a convolutional neural network (CNN) with Deep Supervision based edge detector, which can retain more edge pixels over conventional edge detection algorithms. Initially, the cover image is pre-processed by masking the last 5-bits of each pixel. The said edge detector model is then applied to obtain a gray-scale edge map. To get the prominent edge information, the gray-scale edge map is converted into a binary version using both global and adaptive binarization schemes. The purpose of using different binarization techniques is to prove the less sensitive nature of the edge detection method to the thresholding approaches. Our rule for embedding secret bits within the cover image is as follows: more bits into the edge pixels while fewer bits into the non-edge pixels. Experimental outcomes on various standard images confirm that compared to state-of-the-art methods, the proposed method achieves a higher payload.

     

Pulse-coupled time synchronization for distributed acoustic event detection using wireless sensor networks

Time synchronization has proven to be critical in sensor fusion applications where the time of arrival is utilized as a decision variable. Herein, the application of pulse-coupled synchronization to an acoustic event detection system based on a wireless sensor network is presented. The aim of the system is to locate the source of acoustic events utilizing time of arrival measurements for different formations of the sensor network. A distributed localization algorithm is introduced that solves the problem locally using only a subset of the time of arrival measurements and then fuses the local guesses using averaging consensus techniques. It is shown that the pulse-coupled strategy provides the system with the proper level of synchronization needed to enable accurate localization, even when there exists drift between the internal clocks and the formation is not perfectly maintained. Moreover, the distributed nature of pulse-coupled synchronization allows coordinated synchronization and distributed localization over an infrastructure-free ad-hoc network.