基于半监督学习的跌倒检测系统设计

针对老人跌倒时的复杂运动情况,进行跌倒标注的较难实现,提出了基于Tri-training半监督算法的跌倒检测系统。本系统使用3D加速度传感器采集运动加速度数据,然后对数据进行特征提取与部分样本标注,使用Tri-training算法训练分类器,最后使用训练好的分类器进行跌倒识别。具体的数据采集传感器设计为可穿戴式设备,服务器端使用Java编写了一个服务器的程序实现对数据的分析与处理。实验结果表明：该方法使用了大量无标签数据的信息,有效提高了跌倒识别的准确率。实验结果表明：本系统能够满足老年人在日常生活中的需求,对于一些意外跌倒能够给予及时的检测与报警。     

基于置信域伪标签策略的半监督三维目标检测

当前基于点云的三维目标检测方法很大程度上依赖于大规模高质量的三维标注。为了减少所需标签量,基于SESS网络提出了一种新的三维目标检测方法：基于置信域伪标签策略的半监督三维目标检测。首先设计了一种置信域伪标签策略,将学生网络的输出分成有标签和无标签两部分,有标签部分利用ground truth进行全监督学习,无标签部分基于教师网络的类别和对象预测置信分数,利用一种有效的过滤机制,筛选出高质量教师预测,并转换成相应伪标签,用于监督学生网络无标签部分。其次,设计了一个TransVote模块,通过Transformer机制,增强每个点云与其邻域点之间的相互注意,聚合点云局部特征。在10%标记数据、mAP@0.25下,该算法在ScanNetV2和SUN RGB-D数据集上分别超越了基准线8.63%、6.75%,显著提高了半监督三维目标检测算法的检测精度。     

Curvature-direction measures for 3D feature detection

Using dark channel prior—a kind of statistics of the haze-free outdoor images—to remove haze from a single image input is simple and effective. However, due to the use of soft matting algorithm, the method suffers from massive consumption of both memory and time, which largely limits its scalability for large images. In this paper, we present a hierarchical approach to accelerate dark channel based image dehazing. The core of our approach is a novel, efficient scheme for solving the soft matting problem involved in image dehazing, using adaptively subdivided quadtrees built in image space. Acceleration is achieved by transforming the problem of solving a N-variable linear system required in soft matting, to a problem of solving a much smaller m-variable linear system, where N is the number of pixels and m is the number of the corners in the quadtree. Our approach significantly reduces both space and time cost while still maintains visual fidelity, and largely extends the practicability of dark channel based image dehazing to handle large images.     

3D head tracking for fall detection using a single calibrated camera

The head trajectory is an interesting source of information for behavior recognition and can be very useful for video surveillance applications, especially for fall detection. Consequently, much work has been done to track the head in the 2D image plane using a single camera or in a 3D world using multiple cameras. Tracking the head in real-time with a single camera could be very useful for fall detection. Thus, in this article, an original method to extract the 3D head trajectory of a person in a room is proposed using only one calibrated camera. The head is represented as a 3D ellipsoid, which is tracked with a hierarchical particle filter based on color histograms and shape information. Experiments demonstrated that this method can run in quasi-real-time, providing reasonable 3D errors for a monocular system. Results on fall detection using the head 3D vertical velocity or height obtained from the 3D trajectory are also presented.     

Symmetry-based monocular vehicle detection system

In this paper, we describe the development of a symmetry-based vehicle detection system. The system uses a single forward looking camera to capture the road scene. Vehicles are detected based on their edges and symmetrical characteristics. A method to extract the symmetric regions in the image using a multi-sized window and clustering technique is introduced. We hypothesize the vehicle’s locations in the image from the detected symmetric regions and the regions are then further processed to enhance their symmetrical edges. A bounding box of a vehicle is detected from the projection maps of the enhanced vertical and horizontal edges. The hypothesized vehicles are then verified using a two-class classifier, which consists of an edge oriented histogram (EOH) feature extractor and a support vector machine (SVM). Once a vehicle is verified, a tracking process based on a Kalman filter and a reliability point system is used to track the movement of the vehicle in consecutive video frames. The system was successfully implemented and tested on a standard PC. Experimental results on live video feed and pre-recorded video sequences for various road scenes showed that the system is able to detect multiple vehicles in real time.     

3D map reconstruction using a monocular camera for smart cities

Large-scale high-resolution three-dimensional (3D) maps play a vital role in the development of smart cities. In this work, a novel deep learning-based multi-view-stereo method is proposed for reconstructing the 3D maps in large-scale urban environments by exploiting a monocular camera. Compared with other existing works, the proposed method can perform 3D depth estimation more efficiently in terms of computational complexity and graphics processing unit memory usage. As a result, the proposed method can practically perform depth estimation for each pixel before generating 3D maps for even large-scale scenes. Extensive experiments on the well-known DTU dataset and real-life data collected on our campus confirm the good performance of the proposed method.

     

Testing the reliability of monocular obstacle detection methods in a simulated 3D factory environment

Journal of Intelligent Manufacturing - Automated driving in public traffic still faces many technical and legal challenges. However, automating vehicles at low speeds in controlled industrial...     

A smartphone-based fall detection system

Falls are a major cause of injuries and hospital admissions among elderly people. Thus, the caregiving process and the quality of life of older adults can be improved by adopting systems for the automatic detection of falls. This paper presents a smartphone-based fall detection system that monitors the movements of patients, recognizes a fall, and automatically sends a request for help to the caregivers. To reduce the problem of false alarms, the system includes novel techniques for the recognition of those activities of daily living that could be erroneously mis-detected as falls (such as sitting on a sofa or lying on a bed). To limit the intrusiveness of the system, a small external sensing unit can also be used for the acquisition of movement data.     

3D monocular robotic ball catching

A new method to catch a thrown ball with a robot endowed with an eye-in-hand monocular visual system integrated into a gripper is proposed. As soon as the thrown ball is recognized by the visual system, the camera carried by the robot end-effector is forced to follow a baseline in the space so as to acquire an initial dataset of visual measurements from several points of view, providing a first estimate of the catching point through a linear estimation algorithm. Hereafter, additional measurements are acquired to constantly refine the previous estimate by exploiting a nonlinear estimation algorithm. During the robot trajectory, the translational components of the camera are controlled in such a way as to follow the planned path to intercept the ball, while the rotational components are forced to keep the ball into the field of view. Experimental results performed on a common industrial robotic system prove the effectiveness of the presented solution.     

Shilling attack detection utilizing semi-supervised learning method for collaborative recommender system

Collaborative filtering (CF) technique is capable of generating personalized recommendations. However, the recommender systems utilizing CF as their key algorithms are vulnerable to shilling attacks which insert malicious user profiles into the systems to push or nuke the reputations of targeted items. There are only a small number of labeled users in most of the practical recommender systems, while a large number of users are unlabeled because it is expensive to obtain their identities. In this paper, Semi-SAD, a new semi-supervised learning based shilling attack detection algorithm is proposed to take advantage of both types of data. It first trains a naïve Bayes classifier on a small set of labeled users, and then incorporates unlabeled users with EM-λ to improve the initial naïve Bayes classifier. Experiments on MovieLens datasets are implemented to compare the efficiency of Semi-SAD with supervised learning based detector and unsupervised learning based detector. The results indicate that Semi-SAD can better detect various kinds of shilling attacks than others, especially against obfuscated and hybrid shilling attacks.     

A semi-supervised system for players detection and tracking in multi-camera soccer videos

This paper presents a complete, general and modular system which after a simple previous configuration is able to detect and track each player on the court or field. The presented multi-camera system is based on a mono-camera object detection and tracking system originally designed for video surveillance applications. Target sports of the developed system are team sports (e.g., basketball, soccer). The main objective of this paper is to present a semi-supervised system able to detect and track the players in multi-camera sports videos, focusing on the fusion of different tracks of detected blobs in order to match tracks across cameras. The proposed system is simpler than other systems from the state of the art, can operate in real time and has margin to be improved and to reduce supervision adding additional complexity. In addition to the detection and tracking system, an evaluation system has been designed to obtain quantitative results of the system performance.     

一种半监督三维模型语义自动标注方法

三维模型语义自动标注的目标是自动给出最适合描述模型的标注词集合,是基于文本的三维模型检索的重要环节。语义鸿沟的存在使得相似匹配技术得到的标注效果有待提高。为了在用户提供的有限模型数量和对应的标注词信息下,在自动标注过程中利用大量的未标注样本改善三维模型的标注性能,提出了一种半监督测度学习方法完成三维模型语义自动标注。该方法首先使用基于图的半监督学习方法扩展已标注模型集合,并给出扩展集合中语义标签表征模型的语义置信度,使用改进的相关成分分析方法学习马氏距离度量,依据学习到的距离和语义置信度形成多语义标注策略。在PSB（Princeton Shape Benchmark）数据集上的测试表明,该方法利用了大量未标注样本参与标注过程,取得了比较好的标注效果。     

3-D position sensing using a passive monocular vision system

Passive monocular 3-D position sensing is made possible by a new calibration scheme that relates depth to focus blur through a composite lens and aperture model. The calibration technique enables the recovery of absolute 3-D position coordinates from image coordinates and measured focus blur. A geometric model of the camera's position and orientation in space is used to transform the camera's imaging coordinates into world coordinates. The relationship between the world coordinate system and the screen coordinate system which includes the amount of focus blur, is developed by modeling the camera imaging arrangement. The modeling proceeds first through the perspective view from a pinhole camera located anywhere in space. The camera's lens and aperture system is investigated to find the relationship between depth and focus blur. The aspect ratio of the frame image is considered. Position accuracies comparable to those in stereo based vision systems are possible without the need for solving the difficult point of correspondence problem     

可穿戴式跌倒检测智能系统设计

为提高对老年人跌倒检测的正确率,设计一种可穿戴式跌倒检测系统.研制基于三轴加速度计的跌倒检测设备,给出系统硬件和软件的实现方案;提出基于反向传播(BP)神经网络的跌倒检测算法,将训练好的网络参数植入研制的可穿戴式跌倒检测设备,实现对跌倒的实时检测.实验结果表明:所研制的跌倒检测智能系统能够有效地区分跌倒与非跌倒,正确率达97.37％.     

基于ZigBee的定位和人体跌倒检测系统

提出了一个具有跌倒定位和远程报警功能的无线跌倒监控系统。跌倒监测是基于人体跌倒过程中加速度曲线的变化特性,由三轴加速度传感器和ARM处理器实现;同时介绍了一种新颖的基于ZigBee的定位方法,当监测到人体跌倒时,一个带有跌倒位置的报警信息将立即发送到医护人员的移动手机上。实验结果表明,本系统可以100%监测到人体跌倒,提供精度为1 m的精确定位,使医护人员在5 s内收到跌倒报警信息。     

从视频中恢复三维人脸的实时方法

三维人脸恢复是视觉交互的一个难点问题,提出了一种从视频中实时恢复三维人脸的新方法.该方法利用主动形状模型进行人脸特征点提取和跟踪,确保了三维形状恢复和特征跟踪的有效性和一致性;采用非刚体形状和运动估计方法构建三维形变基,有效地适应人脸形状变化的多样性;采用非线性优化算法估算人脸姿态和三维形变基参数,实现了三维人脸形状和姿态的实时恢复.实验结果表明,该方法不仅能从视频中实时恢复三维人脸模型,而且可有效跟踪人脸各种姿态的变化.     

A multi-camera vision system for fall detection and alarm generation

Abstract: In-house video surveillance can represent an excellent support for people with some difficulties (e.g. elderly or disabled people) living alone and with a limited autonomy. New hardware technologies and in particular digital cameras are now affordable and they have recently gained credit as tools for (semi-)automatically assuring people's safety. In this paper a multi-camera vision system for detecting and tracking people and recognizing dangerous behaviours and events such as a fall is presented. In such a situation a suitable alarm can be sent, e.g. by means of an SMS. A novel technique of warping people's silhouette is proposed to exchange visual information between partially overlapped cameras whenever a camera handover occurs. Finally, a multi-client and multi-threaded transcoding video server delivers live video streams to operators/remote users in order to check the validity of a received alarm. Semantic and event-based transcoding algorithms are used to optimize the bandwidth usage. A two-room setup has been created in our laboratory to test the performance of the overall system and some of the results obtained are reported.     

一种基于半监督学习的工控网络入侵检测方法

随着工控网络的发展以及工业和和信息化的深度融合,针对工业控制系统的攻击行为大幅度增长,对工控企业造成巨大的经济及财产损失。因此,提出一种基于半监督机器学习的入侵检测技术,该技术充分利用工控网络流量标记的特点,结合多种机器学习算法进行实现,并对算法的性能进行了优化。实验证明,该技术可以有效地检测出工控系统网络中的异常流量。     

FRaC: a feature-modeling approach for semi-supervised and unsupervised anomaly detection

Anomaly detection involves identifying rare data instances (anomalies) that come from a different class or distribution than the majority (which are simply called “normal” instances). Given a training set of only normal data, the semi-supervised anomaly detection task is to identify anomalies in the future. Good solutions to this task have applications in fraud and intrusion detection. The unsupervised anomaly detection task is different: Given unlabeled, mostly-normal data, identify the anomalies among them. Many real-world machine learning tasks, including many fraud and intrusion detection tasks, are unsupervised because it is impractical (or impossible) to verify all of the training data. We recently presented FRaC, a new approach for semi-supervised anomaly detection. FRaC is based on using normal instances to build an ensemble of feature models, and then identifying instances that disagree with those models as anomalous. In this paper, we investigate the behavior of FRaC experimentally and explain why FRaC is so successful. We also show that FRaC is a superior approach for the unsupervised as well as the semi-supervised anomaly detection task, compared to well-known state-of-the-art anomaly detection methods, LOF and one-class support vector machines, and to an existing feature-modeling approach.