Effect of task and eccentricity of the target on detection thresholds in mesopic vision: implications for road lighting

OBJECTIVE: The aim of this work is to assess how adding a driving-related task affects the detection of objects in peripheral vision, under mesopic conditions. BACKGROUND: The main index used to assess the quality of road lighting installations refers to simple detection tasks in foveal vision, which raises methodological and practical questions. METHOD: The experimental design consisted of a three-phase experiment. In the first phase, two groups (control and experimental) performed a peripheral detection task (simple task). Based on these results an individual detection threshold was computed for each participant and eccentricity. A tracking task was performed in Phase 2 for both groups (steering a tracking target along a circuit, on a screen). In the third phase, the control group performed the same task as in Phase 2. The experimental group performed a double task, with a tracking (primary) task and a peripheral detection (secondary) task. RESULTS: The data show an effect of the tracking task and eccentricity on peripheral event detection. The tracking task caused detection performance to decrease from 84.2% to 67.5%, p < .001. CONCLUSION: The small target visibility model used in road lighting may be improved, taking into account the effects of task and eccentricity on target detection. APPLICATION: This study supports improved roadway lighting design by guiding consideration of sign eccentricity and task load.     

Task-dependent processing of tables and graphs

In two experiments participants had to detect changes in periodic sinusoidal functions, displayed in either graphic or tabular displays. Graphs had a major advantage over tables when the task required considering configurations of data. Both displays led to similar results when task performance could rely on inspecting individual data points. With graphs almost all participants reported using the optimal method for detecting changes in the function, i.e., they used the method requiring the least effort to perform the task. With tables only about half used the optimal detection method, and these participants showed transfer of learning of detection methods between tasks. Experience in using a detection method led to improved performance if the new task relied on the same method of detection. These findings demonstrate the need to consider task performance methods when determining the relative value of different displays. The set of tasks for which a display is used is likely to affect performance and needs to be analysed as a whole, since methods employed for one task can affect task performance in other tasks.     

信息系统分布式安全检测模型及其设计

提出了一种分布式安全检测模型,可应用不同规模信息系统的安全检测。系统采用攻击和验证相结合的方法,从正反两个方向对被测系统进行检测,并给出了一种检测作业、检测任务和检测状态相分离的作业运行机制,以减少多级检测生成的复杂度。     

基于迁移学习的讽刺检测

准确的讽刺检测对于情感分析等任务至关重要.传统的方法严重依赖于离散的人工制定的特征.现有的研究大多将讽刺检测作为一种标准的监督学习文本分类任务,但是监督学习需要有大量数据,而这些数据的收集和标注都存在困难.由于目标任务有限的数据集可能导致讽刺检测的低性能,为此将讽刺检测作为一种迁移学习任务,将讽刺标记文本的监督学习与外部分析资源的知识转移相结合.通过转移的资源知识来改进神经网络模型,以此提升对目标任务的检测性能.在公开可用的数据集上的实验结果表明,提出的基于迁移学习的讽刺检测模型优于现有较先进的讽刺检测模型.     

面向互联网新闻的在线事件检测

为了提高互联网上新闻事件在线检测的效率,利用加窗策略、命名实体识别及后缀树聚类等技术提出了一种新的检测算法.该算法基于实体识别技术解析出新闻数据特有的信息元素(例如日期、地点、人物等),并在限定的时间窗口内,通过新闻特征的语义匹配实现了新事件的快速识别,从而大幅降低了基于文本相似度计算的检测算法带来的巨大时间消耗.实验结果证明,该算法能够实现在保障检测准确率的同时显著提高检测的效率.     

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.     

改进YOLO框架的血细胞检测算法

为解决传统目标检测算法在血细胞检测任务中出现的检测精度偏低、错检及漏检等问题,提出了一种基于YOLO框架的改进目标检测算法YOLO-Att,该算法在YOLO框架结构的基础上,在骨干网络中增加了一种多尺度残差增强模块,结合低层信息丰富网络的特征层次,进而提高特征信息利用率;并设计了一种注意力门控结构嵌入模型,以获取更多高质量的主要特征信息;同时使用Focal loss代替原损失函数中的交叉熵,提高正负样本权重,加快模型收敛速度;采用[K]-means++聚类算法对目标进行锚框优化,进一步提升检测准确率。相较于现有的Faster-RCNN、SSD以及YOLOv4等目标检测算法,YOLO-Att在通用血细胞数据集BCCD检测任务中,将mAP提高至66.32%,检测速率达到了85.4?ms,更符合血细胞检测任务的实时性。     

基于深度卷积神经网络的航空器检测与识别

针对军用机场大尺寸卫星图像中航空器检测识别的具体应用场景,建立了一套实时目标检测识别框架,将深度卷积神经网络应用到大尺寸图像中的航空器目标检测与识别任务中。首先,将目标检测的任务看成空间上独立的bounding-box的回归问题,用一个24层卷积神经网络模型来完成bounding-box的预测;然后,利用图像分类网络来完成目标切片的分类任务。大尺寸图像上的传统目标检测识别算法通常在时间效率上很难突破,而基于卷积神经网络的航空器目标检测识别算法充分利用了计算硬件的优势,大大缩短了任务耗时。在符合应用场景的自采数据集上进行测试,所提算法目标检测实时性达到平均每张5.765 s,在召回率65.1%的工作点上达到了79.2%的精确率,分类网络的实时性达到平均每张0.972 s,Top-1错误率为13%。所提框架在军用机场大尺寸卫星图像中航空器检测识别的具体应用问题上提出了新的解决思路,同时保证了实时性和算法精度。     

Automatic detection of speaker state: Lexical,prosodic, and phonetic approaches to level-of-interest and intoxication classification

Traditional studies of speaker state focus primarily upon one-stage classification techniques using standard acoustic features. In this article, we investigate multiple novel features and approaches to two recent tasks in speaker state detection: level-of-interest (LOI) detection and intoxication detection. In the task of LOI prediction, we propose a novel Discriminative TFIDF feature to capture important lexical information and a novel Prosodic Event detection approach using AuToBI; we combine these with acoustic features for this task using a new multilevel multistream prediction feedback and similarity-based hierarchical fusion learning approach. Our experimental results outperform published results of all systems in the 2010 Interspeech Paralinguistic Challenge – Affect Subchallenge. In the intoxication detection task, we evaluate the performance of Prosodic Event-based, phone duration-based, phonotactic, and phonetic-spectral based approaches, finding that a combination of the phonotactic and phonetic-spectral approaches achieve significant improvement over the 2011 Interspeech Speaker State Challenge – Intoxication Subchallenge baseline. We discuss our results using these new features and approaches and their implications for future research.     

CRIM’s content-based audio copy detection system for TRECVID 2009

We report results on audio copy detection for TRECVID 2009 copy detection task. This task involves searching for transformed audio queries in over 385?h of test audio. The queries were transformed in seven different ways, three of them involved mixing unrelated speech to the original query, making it a much more difficult task. We give results with two different audio fingerprints and show that mapping each test frame to the nearest query frame (nearest-neighbor fingerprint) results in robust audio copy detection. The most difficult task in TRECVID 2009 was to detect audio copies using predetermined thresholds computed from 2008 data. We show that the nearest-neighbor fingerprints were robust to even this task and gave actual minimal normalized detection cost rate (NDCR) of around 0.06 for all the transformations. These results are close to those obtained by using the optimal threshold for each transform. This result shows the robustness of the nearest-neighbor fingerprints. These nearest-neighbor fingerprints can be efficiently computed on a graphics processing unit, leading to a very fast search.     

基于梯度的重叠式层次社区检测

社区检测(community detection)任务一直是数据挖掘领域的一个研究热点,近年来,深度学习和图链接数据呈现出多样化和复杂化的发展趋势,层次(Hierarchical)社区检测逐渐成为研究的焦点.层次社区检测任务的目标是,在将同质图中相似的节点聚集到社区中的同时,学习社区之间的层次结构关系,以更好的理解图数据结构.社区间层次关系的引入给社区检测算法带来了更复杂的建模挑战.针对该任务,已经有一些有效的启发式的方法被提出,但是受限于社区分布形态的简单假设和离散的优化学习方式,它们无法描述更复杂的图链路数据,也无法和其它有效的连续优化算法组合获得更好的结果.为了解决这个问题,本文首次尝试建模复杂的重叠式(overlapping)层次社区结构,提出简洁的节点嵌入和社区检测双任务优化模型,通过梯度更新的方式来灵活地探索节点和重叠式层次社区的隶属关系.在学习过程中,我们可以分别获得节点和社区的嵌入表示,以应用于丰富的下游任务.     

面向拥挤行人检测的CA-YOLOv5

针对YOLOv5在拥挤行人检测任务中漏检率高、特征融合不充分等问题,提出了CA-YOLOv5行人检测算法。针对原主干网络对细粒度特征融合不充分的问题,采用Res2Block重建YOLOv5的主干网络,以提升网络的细粒度特征融合能力,提高检测精度。针对数据集目标尺度变化大的问题,引入coordinate attention（CA）模块增强感受野,增强模型对目标的精确定位能力。针对FPN结构在特征融合时导致多尺度特征表达能力下降的问题,提出特征增强模块,以增强多尺度特征的表达能力。通过结构重参数化的方法减少模型的计算量与参数量,加快目标检测速度。针对行人检测任务中普遍存在的拥挤行人问题,提出EViT模块,增强模型关注局部信息的能力,提高检测精度。实验证明,在拥挤行人检测任务中,CA-YOLOv5的检测精度达到84.86%,相较于原算法提高了3.75%,检测速度可以达到51?FPS,具有较好的检测精度与实时性。因此,CA-YOLOv5可以更好地应用于实时行人检测任务中。     

A detection method for low-pixel ratio object

Multimedia Tools and Applications - Low-pixel object detection is a kind of difficult program. Existing object detection benchmarks and methods mainly focus on standard detection task. However,...     

基于迁移学习的商品图像检测方法

近年来,对象识别方法被应用到多个领域.如人脸检测,车辆检测.然而模型训练所需要的边框标定需要很大的工作量.本文通过基于迁移学习的方法,将物体检测任务迁移到商品检测,且不需要边框标定.本文在分类层和边框回归层之间建立关系层,来学习两种任务之间的关联.本文建立了一个商品数据集,并提出了一种深度学习训练方法,解决了可旋转物体的检测问题.基于Faster RCNN框架,本文提出一种候选选择方法,可以在无边框标定情况下训练商品分类.本文提出的商品检测方法不需要边框标定,而且很容易训练并应用到其它数据集.     

Image resampling detection based on texture classification

Multimedia Tools and Applications - This study presents a method for resampling detection. By combining texture analysis with resampling detection, the task of resampling detection is considered as...     

基于序贯变化检测的DDoS攻击检测方法

给出了一种有效的DDoS攻击检测方法,将DDoS攻击的检测作为序贯变化检测的一个具体实例来分析,采用序贯变化检测算法--非参数CUSUM算法进行检测.方法具有计算量小、检测迅速准确、适用于不同网络环境和攻击模式的优点,有一定的实用价值.文章最后对两种典型的攻击模式进行了实际检测,全面评估了检测算法在不同DDoS攻击场景下的性能.     

基于Fast R-CNN的车辆目标检测

目的 在传统车辆目标检测问题中,需要针对不同图像场景选择适合的特征。为此提出一种基于快速区域卷积神经网络（Fast R-CNN）的场景图像车辆目标发现方法,避免传统车辆目标检测问题中需要设计手工特征的问题。方法 该方法基于深度学习卷积神经网络思想。首先使用待检测车辆图像定义视觉任务。利用选择性搜索算法获得样本图像的候选区域,将候选区域坐标与视觉任务示例图像一起输入网络学习。示例图像经过深度卷积神经网络中的卷积层,池化层计算,最终得到深度卷积特征。在输入时没有规定示例图像的规格,此时得到的卷积特征规格不定。然后,基于Fast R-CNN网络结构,通过感兴趣区域池化层规格化特征,最后将特征输入不同的全连接分支,并行回归计算特征分类,以及检测框坐标值。经过多次迭代训练,最后得到与指定视觉任务强相关的目标检测模型,具有训练好的权重参数。在新的场景图像中,可以通过该目标检测模型检测给定类型的车辆目标。结果 首先确定视觉任务包含公交车,小汽车两类,背景场景是城市道路。利用与视觉任务强相关的测试样本集对目标检测模型进行测试,实验表明,当测试样本场景与视觉任务相关度越高,且样本中车辆目标的形变越小,得到的车辆目标检测模型对车辆目标检测具有良好的检测效果。结论 本文提出的车辆目标检测方法,利用卷积神经网络提取卷积特征代替传统手工特征提取过程,通过Fast R-CNN对由示例图像组成定义的视觉任务训练得到了效果良好的车辆目标检测模型。该模型可以对与视觉任务强相关新场景图像进行效果良好的车辆目标检测。本文结合深度学习卷积神经网络思想,利用卷积特征替代传统手工特征,避免了传统检测问题中特征选择问题。深层卷积特征具有更好的表达能力。基于Fast R-CNN网络,最终通过多次迭代训练得到车辆检测模型。该检测模型对本文规定的视觉任务有良好的检测效果。本文为解决车辆目标检测问题提供了更加泛化和简洁的解决思路。     

The effect of stimulus modality on signal detection: implications for assessing the safety of in-vehicle technology

OBJECTIVE: This study examined the effect of two in-vehicle information systems (IVIS) on signal detection in the visual, auditory, and tactile modalities; established whether the detrimental effects of an IVIS on driving could be quantified by these detection tasks; and examined the effect of stimulus modality on signal detection. BACKGROUND: The peripheral detection task has been used widely for assessing the effects of an IVIS on driving. However, performance on this task relies on drivers' ability to see a series of LEDs, which can be problematic in field tests (e.g., on sunny days). METHOD: Participants responded to one of three detection tasks during a simulated driving experiment. The effect of IVIS interaction on these detection tasks was also measured. Reduced performance in the detection tasks was assumed to indicate a decline in drivers' ability to handle sudden events in the driving task. RESULTS: Response time to all detection tasks increased by around 200 ms when drivers performed the IVIS tasks, as compared with baseline driving. Analyses of variance and comparison of effect sizes showed the effects of these two IVISs to be the same across the three detection tasks. CONCLUSION: These detection tasks are useful for quantifying the safety of an IVIS during driving. The absence of a difference in signal detection by modality suggests that performance on these tasks relies on general attentional resources and is not modality specific. APPLICATION: The signal detection tasks employed here should be further investigated for their suitability in assessing the safety of in-vehicle systems.     

Remarks on Colquhoun's: “ The Effect of ‘ unwanted ’ Signals on Performance in a Vigilance Task ”

Colquhoun's 1961 experiment is interpreted as emphasizing the discrimination phase of a complex task in which the detection of a signal is followed by the discrimination of one of its features. His conclusion that signal probability determines vigilance performance is supported by research from this laboratory with a simple detection task. In our research the probability is redefined as the ratio of signals to attention-eliciting stimuli, and the latter stimuli must be presented at high rates (15 or more times per minute). His results with the complex task suggest an effect of signal probability on search and scanning patterns during prolonged visual work.     

After-effects of TFT-LCD display polarity and display colour on the detection of low-contrast objects

Participants performed a word–non-word discrimination task within a car control display emulated on a thin film transistor liquid-crystal display (TFT-LCD). The task simulated an information read-out from a TFT-LCD-based instrument panel. Subsequently, participants performed a low-contrast object detection task that simulated the detection of objects during night-time driving. In experiment 1, words/non-words were presented black-on-white (positive polarity) or white-on-black (negative polarity). In experiments 2 and 3, display colour was additionally manipulated. A positive polarity advantage in the discrimination task was consistently observed. In contrast, positive displays interfered more than negative displays with subsequent detection. The detrimental after-effect of positive polarity displays was strong with white and blue, reduced with amber and absent with red displays. Subjective measures showed a preference for blue over red, but a slight advantage for amber over blue. Implications for TFT-LCD design are derived from the results.

Statement of Relevance: When using TFT-LCDs as car instrument panels, positive polarity red TFT-LCDs are very likely to lead to good instrument readability while at the same time minimising – relative to other colours – the negative effects of an illuminated display on low-contrast object detection during night-time driving.