SAR图像中目标的自动检测与辨别

SAR图像自动目标识别(automatic target recognition,ATR)是当前的一大研究热点。典型的ATR系统分为检测、辨别和识别3个阶段。在检测和辨别过程中对图像进行预筛选,找出感兴趣区域,是进行目标识别前的一个重要步骤。高效的预筛选过程可以大大减少目标识别过程的计算量。目前,目标检测的方法有CFAR方法、多分辨率方法以及基于相位信息的检测方法3类。目标辨别的方法也有多种。本文就目标检测和辨别阶段的主要算法及其效果作了系统的介绍,并对该领域未来的发展方向进行了展望。     

False alarm moderation for performance monitoring in industrial water distribution systems

While considerable attention has been given to data driven methods that analyse and control energy systems in buildings, the same cannot be said for building water systems. As a result, approaches which support enhanced efficiency in building water consumption are somewhat underdeveloped, particularly in industrial settings. Water consumption in industrial systems features non-stationarity (i.e., variations in statistical properties over time), making it challenging to distinguish between routine and non-routine water uses. In such scenarios, fault detection and diagnosis methods that leverage multivariate statistical process control with, for example, principal component analysis and detection indices (Hotelling T²-statistics and Q-statistics), can be successfully used to identify system alarms. However, even with these approaches there can be a high prevalence of false alarms leading to low industry uptake of fault detection and diagnosis systems, or where in place, alarms can be ignored. To efficiently detect and diagnose water distribution system faults, false alarms should be controlled through false alarm moderation approaches so that building managers/operators only need to focus on critical system alarms or system alarms with high risk levels. This paper utilises two statistical non-parametric false alarm moderation approaches (window-based, and trial-based) that generate a second control limit for T²-statistics and Q-statistics. The implementation of these false alarm moderation approaches was combined with principal component analysis to detect faults with real water time series data from two case-study sites. Using both approaches false alarms were reduced, and the overall performance and reliability of the fault detection and diagnosis approach was improved. The principal component analysis model with the window-based approach was shown to be particularly effective.     

The impact of false alarm mitigation on surface landmine detection in MWIR imagery

Surface landmine and minefield detection from airborne imagery is a difficult problem. As part of the minefield detection process, anomaly detection is performed to identify potential landmines in individual airborne images. Post-processing is performed on the initial landmines identified to reduce the number of false alarms, referred to as false alarm mitigation. In this research, a circular harmonics transform image processing approach (the CHT method) and a constant false alarm rate technique (the RX approach) are investigated for surface landmine detection and false alarm mitigation in medium wave infrared (MWIR) image data. The false alarm mitigation approach integrates the CHT and RX methods to identify candidate landmine locations with one technique at a given false alarm rate and applies the other technique to confirm landmine locations and eliminate potential false alarms. Individual detector and false alarm mitigation experimental results are presented for 31 daytime and 43 nighttime MWIR images containing 76 and 142 landmines, respectively. At a 0.9 desired probability of landmine detection, experimental results show that false alarm mitigation reduces the false alarm rate by as much as 84.3% and 13.7% for daytime and nighttime images, respectively, maintaining the probability of detection at 0.85 and 0.90, respectively.     

基于累积量和主运动方向的视频烟雾检测方法

视频烟雾检测具有响应速度快、非接触等优点。但现有的视频检测方法误报率比较高。通过分析早期火灾烟雾运动规律,提出了一种适用于普通视频的烟雾检测方法。为了加快检测速度,将视频图像分割成大小相等的块,并估计每个块的运动方向。采用滑动时间窗口生成块运动方向时间序列,在此时间序列的基础上计算块的累积量和主运动方向。累积量可以反映出运动持续的程度,而主运动方向表明每个块最可能的运动方向,可以有效地抑制噪声的干扰。根据累积量和主运动方向提取出3维特征矢量,采用贝叶斯分类器进行烟雾的检测。实验结果表明,该方法鲁棒性高、速度快,能够准确地检测烟雾的出现。     

Human Face Detection in Digital Video Using SVMEnsemble

This Letter proposes automatic human face detection in digital video using a support vector machine (SVM) ensemble to improve the detection performance. The SVM ensemble consists of several independently trained SVMs using randomly chosen training samples via a bootstrap technique. Next, they are aggregated in order to make a collective decision via a majority voting scheme. Experimental results show that the proposed face detection method using SVM ensemble outperforms conventional methods such as using only single SVM and Multi-Layer Perceptron in terms of classification accuracy, false alarms, and missing rates.     

Motion compensated film restoration

Motion picture films are susceptible to local degradations such as dust spots. Other deteriorations are global such as intensity and spatial jitter. It is obvious that motion needs to be compensated for before the detection/correction of such local and dynamic defects. Therefore, we propose a hierarchical motion estimation method ideally suited for high resolution film sequences. This recursive block-based motion estimator relies on an adaptive search strategy and Radon projections to improve processing speed. The localization of dust particles then becomes straightforward. Thus, it is achieved by simple inter-frame differences between the current image and motion compensated successive and preceding frames. However, the detection of spatial and intensity jitter requires a specific process taking advantage of the high temporal correlation in the image sequence. In this paper, we present our motion compensation-based algorithms for removing dust spots, spatial and intensity jitter in degraded motion pictures. Experimental results are presented showing the usefulness of our motion estimator for film restoration at reasonable computational costs. Received: 9 July 2000 / Accepted: 13 January 2002 Correspondence to:S. Boukir     

基于最弱前置条件的静态分析误报消除技术

针对程序静态分析技术误报过多的问题,提出一种基于最弱前置条件的静态分析误报消除方法。根据不同的软件安全性质,从目标状态出发,以需求驱动的方式得到过程起始位置的最弱前置条件,判断该条件公式的可满足性来消除误报。将该方法实例化来消除静态分析工具检测数组访问越界和空指针解引用的误报,实验结果表明该方法是有效且实用的。     

Development of neural network committee machines for automatic forest fire detection using lidar

Lidar has considerable potential as an early forest fire detection technique, presenting considerable advantages when compared to the passive detection methods based on infrared cameras currently in common use, due to its higher sensitivity, ability to accurately locate the fire and the fact that it does not need line of sight to the flames. The method has recently been demonstrated by the authors, but its automation requires the availability of a rapid signal analysis technique, for prompt alarm emission whenever required. In the present paper a novel method of classifying lidar signals using committee machines composed of neural networks is proposed. A new method based on ROC curves and the Neyman-Pearson criterion is used to choose the optimal number of training epochs for each neural network in order to avoid overfitting. The best committee machine, obtained on the basis of these principles and selected to lead to the lowest percentage of false alarms for a true detection percentage of 90% for a test set created by adding random noise to patterns obtained experimentally, was composed of three single-layer perceptrons and presented a true detection efficiency of 94.4% and 0.553% of false alarms in the validation set.     

Real-time motion detection based on SW/HW-codesign for walking rescue robots

In a rescue operation walking robots offer a great deal of flexibility in traversing uneven terrain in an uncontrolled environment. For such a rescue robot, each motion is a potential vital sign and the robot should be sensitive enough to detect such motion, at the same time maintaining high accuracy to avoid false alarms. However, the existing techniques for motion detection have severe limitations in dealing with strong levels of ego-motion on walking robots. This paper proposes an optical flow-based method for the detection of moving objects using a single camera mounted on a hexapod robot. The proposed algorithm estimates and compensates ego-motion to allow for object detection from a continuously moving robot, using a first-order-flow motion model. Our algorithm can deal with strong rotation and translation in 3D, with four degrees of freedom. Two alternative object detection methods using a 2D-histogram based vector clustering and motion-compensated frame differencing, respectively, are examined for the detection of slow- and fast-moving objects. The FPGA implementation with optimized resource utilization using SW/HW codesign can process video frames in real-time at 31 fps. The new algorithm offers a significant improvement in performance over the state-of-the-art, under harsh environment and performs equally well under smooth motion.     

一种基于Cascade R-CNN的电子器件容器质检方法

电子器件容器生产是一种对安全性、高效性、完整性要求极高的过程,是各大企业必须要关注的问题。但是在实际的生产封装过程中,容器上的污渍、容器内的异物,外观的异常不可避免地出现,这些问题亟待解决。目前解决这些问题主要的检测方法还是人工检测和传统的机器视觉的方式,人工检测方式的缺点在于准确率高而效率低,传统机器视觉检测方式是效率高而准确率低,都难以满足高速自动化生产线要求。因此,本文提出一种基于Cascade R-CNN的电子器件容器质检方法,针对实际过程中的容器数据定向改进网络,加入Focal Loss检测难以区分的样本,使用可变形卷积更高效地提取特征,以多尺度训练方式训练强鲁棒性的模型,用于电子器件容器的多类别检测问题。实验结果表明提出的改进的基于Cascade R-CNN的电子器件容器质检模型具有高准确率和强鲁棒性。     

基于卷积神经网络的雷达目标检测方法

雷达目标检测近年来一直是雷达信号处理中的重要任务,在探测监控等安全领域中有非常重要的作用;针对传统恒虚警目标检测方法存在的环境适应能力较弱、复杂地形环境下雷达虚警数量急剧上升等问题,提出一种基于卷积神经网络的雷达目标检测方法;以雷达回波信号数据处理后得到的距离-多普勒图像作为模型的训练集和测试集,设计基于FasterR-CNN结构的雷达目标检测模型,训练模型并将测试结果与传统恒虚警目标检测算法结果相比较,所设计的模型提升了雷达目标检测正确率并较大地减少了虚警数量,这表明将卷积神经网络应用于雷达回波信号的处理任务中是可行的。     

Crowdsourced reliable labeling of safety-rule violations on images of complex construction scenes for advanced vision-based workplace safety

Construction workplace hazard detection requires engineers to analyze scenes manually against many safety rules, which is time-consuming, labor-intensive, and error-prone. Computer vision algorithms are yet to achieve reliable discrimination of anomalous and benign object relations underpinning safety violation detections. Recently developed deep learning-based computer vision algorithms need tens of thousands of images, including labels of the safety rules violated, in order to train deep-learning networks for acquiring spatiotemporal reasoning capacity in complex workplaces. Such training processes need human experts to label images and indicate whether the relationship between the worker, resource, and equipment in the scenes violate spatiotemporal arrangement rules for safe and productive operations. False alarms in those manual labels (labeling no-violation images as having violations) can significantly mislead the machine learning process and result in computer vision models that produce inaccurate hazard detections. Compared with false alarms, another type of mislabels, false negatives (labeling images having violations as “no violations”), seem to have fewer impacts on the reliability of the trained computer vision models.This paper examines a new crowdsourcing approach that achieves above 95% accuracy in labeling images of complex construction scenes having safety-rule violations, with a focus on minimizing false alarms while keeping acceptable rates of false negatives. The development and testing of this new crowdsourcing approach examine two fundamental questions: (1) How to characterize the impacts of a short safety-rule training process on the labeling accuracy of non-professional image annotators? And (2) How to properly aggregate the image labels contributed by ordinary people to filter out false alarms while keeping an acceptable false negative rate? In designing short training sessions for online image annotators, the research team split a large number of safety rules into smaller sets of six. An online image annotator learns six safety rules randomly assigned to him or her, and then labels workplace images as “no violation” or ‘violation” of certain rules among the six learned by him or her. About one hundred and twenty anonymous image annotators participated in the data collection. Finally, a Bayesian-network-based crowd consensus model aggregated these labels from annotators to obtain safety-rule violation labeling results. Experiment results show that the proposed model can achieve close to 0% false alarm rates while keeping the false negative rate below 10%. Such image labeling performance outdoes existing crowdsourcing approaches that use majority votes for aggregating crowdsourced labels. Given these findings, the presented crowdsourcing approach sheds lights on effective construction safety surveillance by integrating human risk recognition capabilities into advanced computer vision.     

基于运动累积和半透明的视频烟雾探测模型

通过分析早期火灾烟雾运动规律,提出了一种适用于普通可见光视频的运动累积和半透明的视频烟雾探测模型。由于烟雾通常从阴燃点持续冒出,因而通过累积模型度量运动像素的累积程度,能够很好地捕获这种早期火灾的时空视觉特征,同时有效地抑制噪声的干扰。根据烟雾的模糊和部分遮挡背景特性,提出了一种基于高通滤波的半透明遮挡快速模型。半透明模型能有效地表征烟雾遮挡的半透明特性。实验结果表明,累积和半透明模型相结合提高了鲁棒性、增强了抗干扰性,明显地提高了探测的准确率。     

A framework for interactive work design based on motion tracking,simulation, and analysis

Because of the flexibility and adaptability of humans, manual handling work is still important in industry, especially in assembly and maintenance work. Well‐designed work operation can improve work efficiency and quality; enhance safety, and lower cost. Most traditional methods for work system analysis need physical mock‐ups and are time‐consuming. Digital mock‐up (DMU) and digital human modeling (DHM) techniques have been developed to assist ergonomic design and evaluation for a specific worker population (e.g., 95 percentile); however, the operation adaptability and adjustability for a specific individual are not considered enough. In this study, a new framework based on motion‐tracking technique and digital human simulation technique is proposed for motion–time analysis of manual operations. A motion‐tracking system is used to track a worker's operation while he/she is conducting a manual handling task. The motion data are transferred to a simulation computer for real‐time digital human simulation. The data are also used for motion type recognition and analysis either online or offline for objective work efficiency evaluation and subjective work task evaluation. Methods for automatic motion recognition and analysis are presented. Constraints and limitations of the proposed method are discussed. © 2010 Wiley Periodicals, Inc.     

A novel approach to detecting DDoS Attacks at an Early Stage

Distributed Denial-of-Service (DDoS) attacks pose a serious threat to Internet security. Most current research focuses on detection and prevention methods on the victim server or source side. To date, there has been no work on defenses using valuable information from the innocent client whose IP has been used in attacking packets. In this paper, we propose a novel cooperative system for producing warning of a DDoS attack. The system consists of a client detector and a server detector. The client detector is placed on the innocent client side and uses a Bloom filter-based detection scheme to generate accurate detection results yet consumes minimal storage and computational resources. The server detector can actively assist the warning process by sending requests to innocent hosts. Simulation results show that the cooperative technique presented in this paper can yield accurate DDoS alarms at an early stage. We theoretically show the false alarm probability of the detection scheme, which is insensitive to false alarms when using specially designed evaluation functions. This work is partially supported by HK Polyu ICRG A-PF86 and CERG Polyu 5196/04E, and by the National Natural Science Foundation of China under Grant No. 90104005.     

Robust scale invariant target detection using the scale-space theory and optimization for IRST

This paper presents a new small target detection method using scale invariant feature. Detecting small targets whose sizes are varying is very important to automatic target detection in infrared search and track (IRST). The conventional spatial filtering methods with fixed sized kernel show limited target detection performance for incoming targets. The scale invariant target detection can be defined as searching for maxima in the 3D (x, y, and scale) representation of an image with the Laplacian function. The scale invariant feature can detect different sizes of targets robustly. Experimental results with real FLIR images show higher detection rate and lower false alarm rate than conventional methods. Furthermore, the proposed method shows very low false alarms in scan-based IR images than conventional filters.     

Improving flood monitoring by the Robust AVHRR Technique (RAT) approach: the case of the April 2000 Hungary flood

In the past, satellite remote sensing techniques have been widely used within the flood risk management cycle. In particular, there have been many demonstrations of the operational use of satellite data for detailed monitoring and mapping of floods and for post-flood damage assessment. When frequent situation reports are requested (e.g. in the emergency phase or for early warning purposes) to assist civil protection activities, high temporal resolution satellites (mainly meteorological, with revisiting times from hours to minutes) can play a strategic role. In this paper, a new Advanced Very High Resolution Radiometer (AVHRR) technique for monitoring flooded areas is presented. Its performances are evaluated in comparison with other well-known approaches, analysing the flood event that occurred in Hungary during April 2000 involving the Tisza and Timis Rivers. The preliminary results seem to indicate the benefits of such a new technique, especially when different observational conditions are considered. In fact, compared with previously proposed techniques, the proposed approach: (a) is completely automatic (i.e. unsupervised with no need for operator intervention); (b) improves flooded-area detection capabilities strongly reducing false alarms; and (c) automatically discriminates (without the need for ancillary information) flooded areas from permanent water bodies. Moreover, it is globally applicable and, because of the complete independence on the specific satellite platform, is easily exportable to different satellite packages.     

A real-time object detecting and tracking system for outdoor night surveillance

Autonomous video surveillance and monitoring has a rich history. Many deployed systems are able to reliably track human motion in indoor and controlled outdoor environments. However, object detection and tracking at night remain very important problems for visual surveillance. The objects are often distant, small and their signatures have low contrast against the background. Traditional methods based on the analysis of the difference between successive frames and a background frame will do not work. In this paper, a novel real time object detection algorithm is proposed for night-time visual surveillance. The algorithm is based on contrast analysis. In the first stage, the contrast in local change over time is used to detect potential moving objects. Then motion prediction and spatial nearest neighbor data association are used to suppress false alarms. Experiments on real scenes show that the algorithm is effective for night-time object detection and tracking.