Irrelevant Objects Research Articles

Anticipatory Distractor Suppression Elicited by Statistical Regularities in Visual Search.

Salient yet irrelevant objects often capture our attention and interfere with our daily tasks. Distraction by salient objects can be reduced by suppressing the location where they are likely to appear. The question we addressed here was whether suppression of frequent distractor locations is already implemented beforehand, in anticipation of the stimulus. Using EEG, we recorded cortical activity of human participants searching for a target while ignoring a salient distractor. The distractor was presented more often at one location than at any other location. We found reduced capture for distractors at frequent locations, indicating that participants learned to avoid distraction. Critically, we found evidence for "proactive suppression" as already "prior to display onset," there was enhanced power in parieto-occipital alpha oscillations contralateral to the frequent distractor location-a signal known to occur in anticipation of irrelevant information. Locked to display onset, ERP analysis showed a distractor suppression-related distractor positivity (PD) component for this location. Importantly, this PD was found regardless of whether distracting information was presented at the frequent location. In addition, there was an early PD component representing an early attentional index of the frequent distractor location. Our results show anticipatory (proactive) suppression of frequent distractor locations in visual search already starting prior to display onset.

SDFN: Segmentation-based deep fusion network for thoracic disease classification in chest X-ray images

This study aims to automatically diagnose thoracic diseases depicted on the chest x-ray (CXR) images using deep convolutional neural networks. The existing methods generally used the entire CXR images for training purposes, but this strategy may suffer from two drawbacks. First, potential misalignment or the existence of irrelevant objects in the entire CXR images may cause unnecessary noise and thus limit the network performance. Second, the relatively low image resolution caused by the resizing operation, which is a common pre-processing procedure for training neural networks, may lead to the loss of image details, making it difficult to detect pathologies with small lesion regions. To address these issues, we present a novel method termed as segmentation-based deep fusion network (SDFN), which leverages the domain knowledge and the higher-resolution information of local lung regions. Specifically, the local lung regions were identified and cropped by the Lung Region Generator (LRG). Two CNN-based classification models were then used as feature extractors to obtain the discriminative features of the entire CXR images and the cropped lung region images. Lastly, the obtained features were fused by the feature fusion module for disease classification. Evaluated by the NIH benchmark split on the Chest X-ray 14 Dataset, our experimental result demonstrated that the developed method achieved more accurate disease classification compared with the available approaches via the receiver operating characteristic (ROC) analyses. It was also found that the SDFN could localize the lesion regions more precisely as compared to the traditional method.

Floating Xylene Spill Segmentation from Ultraviolet Images via Target Enhancement

Automatic colorless floating hazardous and noxious substances (HNS) spill segmentation is an emerging research topic. Xylene is one of the priority HNSs since it poses a high risk of being involved in an HNS incident. This paper presents a novel algorithm for the target enhancement of xylene spills and their segmentation in ultraviolet (UV) images. To improve the contrast between targets and backgrounds (waves, sun reflections, and shadows), we developed a global background suppression (GBS) method to remove the irrelevant objects from the background, which is followed by an adaptive target enhancement (ATE) method to enhance the target. Based on the histogram information of the processed image, we designed an automatic algorithm to calculate the optimal number of clusters, which is usually manually determined in traditional cluster segmentation methods. In addition, necessary pre-segmentation processing and post-segmentation processing were adopted in order to improve the performance. Experimental results on our UV image datasets demonstrated that the proposed method can achieve good segmentation results for chemical spills from different backgrounds, especially for images with strong waves, uneven intensities, and low contrast.

Video instance search via spatial fusion of visual words and object proposals

Most popular systems for object instance search are based on the bag-of-visual-word model. The inherent weaknesses of this standard model such as quantization error, unstructured representation, burstiness phenomenon are to some extent solved. However, it has a serious problem of searching small objects on a database with cluttered background. In many situations, even the irrelevant objects which share the same texture or shape with a query object get higher score than relevant ones. To overcome this problem, we propose a novel fusion method to significantly boost the accuracy of instance search systems. Firstly, we use the state-of-the-art object detector with denser feature for finding object bounding box and similarity score. Secondly, to exploit the spatial relationship of each visual word with an object proposal, a detected area that might contain a query object, we define three categories of visual word pairs, i.e., discriminative, weak relevant, and context inferred ones. Finally, we propose a new re-ranking scheme with three weighting functions corresponding to the three categories of visual word pairs to compute the final similarity score between a query topic and a video shot. To illustrate the efficiency of the proposed method, we conduct experiments on datasets which have a wide variety of types of query objects. Experimental results on TRECVID Instance Search datasets (INS2013 and INS2014) show the superiority of our proposed method over the state-of-the-art approaches.

Grasping and perception are both affected by irrelevant information and secondary tasks: new evidence from the Garner paradigm.

In their Perception-Action Model (PAM), Goodale and Milner (1992) proposed functionally independent and encapsulated processing of visual information for action and perception. In this context, they postulated that visual input for action is processed in an automatized and analytic manner, which renders visuomotor behaviour immune to perceptual interferences or multitasking costs due to sharing of cognitive resources. Here, we investigate the well-known Garner Interference effect under dual- and single-task conditions in its classic perceptual form as well as in grasping. Garner Interference arises when stimuli are classified along a relevant dimension (e.g., their length), while another irrelevant dimension (e.g., their width) has to be ignored. In the present study, participants were presented with differently sized rectangular objects and either grasped them or classified them as long or short via button presses. We found classical Garner Interference effects in perception as expressed in prolonged reaction times when variations occurred also in the irrelevant object dimension. While reaction times during grasping were not susceptible to Garner Interference, effects were observed in a number of measures that reflect grasping accuracy (i.e., poorer adjustment of grip aperture to object size, prolonged adjustment times, and increased variability of the maximum hand opening when irrelevant object dimensions were varied). In addition, multitasking costs occurred in both perception and action tasks. Thus, our findings challenge the assumption of automaticity in visuomotor behaviour as proposed by the PAM.

Creating 3D city models with building footprints and LIDAR point cloud classification: A machine learning approach

The increasing availability of open geospatial data, such as building footprint vector data and LiDAR (Light Detection and Ranging) point clouds, has provided opportunities to generate large-scale 3D city models at low cost. However, using unclassified point clouds with building footprints to estimate building heights may yield erroneous results due to potential errors and anomalies in both datasets and their integration. Some of the points within footprints often reflect irrelevant objects other than roofs, leading to biases in height estimation, and few studies have developed systematic methods to filter them out. In this paper, a LiDAR point classification methodology is proposed that extracts only rooftop points for building height estimation. The LiDAR points are characterized by point, footprint, and neighborhood-based features and classified by the Random Forest (RF) algorithm into four classes – rooftop, wall, ground, and high outlier. The percentage of correctly classified points among 15,577 sample points in Columbus, Ohio, amounts to 96.5%. Conducting this classification separately for different building types (commercial, residential, skyscraper, and small constructions) does not significantly change the overall accuracy. The footprint-based features contribute most to predicting the classes correctly. Height validation results based on a sample of 498 buildings show that (1) using average or median heights with classified points provides the best estimates, minimizing the disparities between computed heights and ground truth and (2) the RF method yields outcomes much closer to ground truth than earlier classification approaches. Some outcomes are visualized in 3D format using Google Earth 3D Imagery and ArcScene.

An Outlier Fuzzy Detection Method Using Fuzzy Set Theory

Outlier mining task is to discover some unusual objects, and however, most existing methods and their mining results lack pertinence. To address the pertinence of outlier results, we propose a novel outlier detection approach, namely, FOD, which aims at finding anomalies in full dimensions that lack pertinence and understandability. Our key idea is to use fuzzy constraint technology to prune irrelevant objects for outlier detection, during which the nearness measure theory in fuzzy mathematics is used for detecting similarities between objects and constraint information. FOD finds outlier by searching sparse subspace, where genetic algorithms can be extended and incorporated into FOD such that an optimum solution of an anomaly is discovered. While constructing a sparse subspace, we present the sparse threshold concept to describe the sparse levels of data objects in a subspace, where data objects are regarded as anomalies. Then, we demonstrate the effectiveness and scalability of our method on synthetic and UCI datasets. The experiment evaluations reveal that our fuzzy constraint-based outlier detection is superior to two existing full dimensional algorithms. Moreover, FOD algorithm also improves the accuracy of outlier detection.

Current and future goals are represented in opposite patterns in object-selective cortex.

Adaptive behavior requires the separation of current from future goals in working memory. We used fMRI of object-selective cortex to determine the representational (dis)similarities of memory representations serving current and prospective perceptual tasks. Participants remembered an object drawn from three possible categories as the target for one of two consecutive visual search tasks. A cue indicated whether the target object should be looked for first (currently relevant), second (prospectively relevant), or if it could be forgotten (irrelevant). Prior to the first search, representations of current, prospective and irrelevant objects were similar, with strongest decoding for current representations compared to prospective (Experiment 1) and irrelevant (Experiment 2). Remarkably, during the first search, prospective representations could also be decoded, but revealed anti-correlated voxel patterns compared to currently relevant representations of the same category. We propose that the brain separates current from prospective memories within the same neuronal ensembles through opposite representational patterns.

Robust long-term correlation tracking using convolutional features and detection proposals

Abstract Correlation filter based trackers have achieved appealing performance and high efficiency in recent years. However, for long-term tracking where target objects undergo dramatic appearance variation due to heavy occlusion or out-of-view, conventional correlation filter based tracking algorithms would be distracted by irrelevant objects. Once the trained tracker loses its way, it is impossible to recover the information for the following frames as the model has drifted. In this paper, we decompose the long-term tracking task into tracking and detection. Tracker learns separate correlation filters for explicit translation and scale estimation. Specifically, in order to improve tracking accuracy, the convolutional features for translation filter are extracted, and the scale filter is learned using the target appearance sampled at different scales. Detector trains an online long-term filter and applies it to the entire frame to generate detection proposals. By exploiting these detection proposals, it helps the tracker to recover from problems such as temporary or persistent occlusions. In this way, the proposed approach could handle the model drifting problem effectively for long-term tracking with more accurate estimation of object scale and location. Extensive experimental results on large-scale benchmark sequences have shown the robustness of the proposed method.

Building Detection from VHR Remote Sensing Imagery Based on the Morphological Building Index

Automatic detection of buildings from very high resolution (VHR) satellite images is a current research hotspot in remote sensing and computer vision. However, many irrelevant objects with similar spectral characteristics to buildings will cause a large amount of interference to the detection of buildings, thus making the accurate detection of buildings still a challenging task, especially for images captured in complex environments. Therefore, it is crucial to develop a method that can effectively eliminate these interferences and accurately detect buildings from complex image scenes. To this end, a new building detection method based on the morphological building index (MBI) is proposed in this study. First, the local feature points are detected from the VHR remote sensing imagery and they are optimized by the saliency index proposed in this study. Second, a voting matrix is calculated based on these optimized local feature points to extract built-up areas. Finally, buildings are detected from the extracted built-up areas using the MBI algorithm. Experiments confirm that our proposed method can effectively and accurately detect buildings in VHR remote sensing images captured in complex environments.

The influence of irrelevant dangerous objects on the action activation of target objects

Previous studies have shown that object observation can automatically evoke motor programs for specific actions; this is referred to as affordance. Recent studies have evidenced that individuals are sensitive to the difference between dangerous and neutral objects. This study explored the influence of irrelevant dangerous object on individuals’ responses to the target object. In a priming task, participants were required to respond to a target object, which was preceded by a prime object. The object types varied between dangerous and neutral objects. The results revealed that the affordance of the neutral object only occurred when the orientation of the prime object corresponded to that of the target object. The affordance of dangerous object was modulated by a dangerous prime object. When the orientation of the prime object did not correspond to that of the target object, the affordance of the dangerous object was observed; whereas, when the orientation was in correspondence, the congruency effect was reversed. These findings suggest that object-action priming is modulated by the danger level of objects.

Lane Detection and Classification for Forward Collision Warning System Based on Stereo Vision

This paper presents a lightweight stereo vision-based driving lane detection and classification system to achieve the ego-car's lateral positioning and forward collision warning to aid advanced driver assistance systems (ADAS). For lane detection, we design a self-adaptive traffic lanes model in Hough Space with a maximum likelihood angle and dynamic pole detection region of interests (ROIs), which is robust to road bumpiness, lane structure changing while the ego-car's driving and interferential markings on the ground. What's more, this model can be improved with geographic information system or electronic map to achieve more accurate results. Besides, the 3-D information acquired by stereo matching is used to generate an obstacle mask to reduce irrelevant objects' interfere and detect forward collision distance. For lane classification, a convolutional neural network is trained by using manually labeled ROI from KITTI data set to classify the left/right-side line of host lane so that we can provide significant information for lane changing strategy making in ADAS. Quantitative experimental evaluation shows good true positive rate on lane detection and classification with a real-time (15Hz) working speed. Experimental results also demonstrate a certain level of system robustness on variation of the environment.

ROAD DAMAGE EXTRACTION FROM POST-EARTHQUAKE UAV IMAGES ASSISTED BY VECTOR DATA

Abstract. Extraction of road damage information after earthquake has been regarded as urgent mission. To collect information about stricken areas, Unmanned Aerial Vehicle can be used to obtain images rapidly. This paper put forward a novel method to detect road damage and bring forward a coefficient to assess road accessibility. With the assistance of vector road data, image data of the Jiuzhaigou Ms7.0 Earthquake is tested. In the first, the image is clipped according to vector buffer. Then a large-scale segmentation is applied to remove irrelevant objects. Thirdly, statistics of road features are analysed, and damage information is extracted. Combining with the on-filed investigation, the extraction result is effective.

The Closing-In Phenomenon in an Ecological Walking Task.

Alzheimer's disease (AD) patients may show the Closing-in (CI), a tendency to reproduce figures close to or superimposed on the model. AD patients with CI might manifest reduced functional independence compared to AD patients without CI, but no study directly assessed if CI can hamper common daily living activities. To address this issue here we investigated whether AD patients with CI veer their walking trajectory toward irrelevant objects more often than AD patients without CI. Fifty AD individuals, and 20 age- and education-matched healthy adults, underwent a graphic copying task to detect CI and a newly developed walking task to assess the tendency to veer toward irrelevant objects and to bump into them. All participants also completed a comprehensive neuropsychological battery to assess dementia severity; impairments in frontal/executive, visuo-spatial, visuo-constructional, and memory domains; and functional independence in daily living activities. Graphic CI occurred in 34/50 (68%) AD patients (AD-CI group) who achieved significantly lower scores on frontal/executive abilities, and daily living functioning than AD individuals not showing CI. Most AD-CI patients (20/34; 58.8%) also showed at least one veering error in the walking task. Participants with CI and veering errors showed significantly poorer performance on Stroop test, and lower level of functional independence than AD individuals with CI in isolation. CI on graphic tasks can identify difficulties in walking and in complying with everyday activities in AD patients. These observations demonstrate the value of assessing CI in copying tasks. (JINS, 2018, 24, 437-444).

Multiple-Instance Ordinal Regression

Ordinal regression (OR) is a paradigm in supervised learning, which aims at learning a prediction model for ordered classes. The existing studies mainly focus on single-instance OR, and the multi-instance OR problem has not been explicitly addressed. In many real-world applications, considering the OR problem from a multiple-instance aspect can yield better classification performance than from a single-instance aspect. For example, in image retrieval, an image may contain multiple and possibly heterogeneous objects. The user is usually interested in only a small part of the objects. If we represent the whole image as a global feature vector, the useful information from the targeted objects that the user is of interest may be overridden by the noisy information from irrelevant objects. However, this problem fits in the multiple-instance setting well. Each image is considered as a bag, and each object region is treated as an instance. The image is considered as of the user interest if it contains at least one targeted object region. In this paper, we address the multi-instance OR where the OR classifier is learned on multiple-instance data, instead of single-instance data. To solve this problem, we present a novel multiple-instance ordinal regression (MIOR) method. In MIOR, a set of parallel hyperplanes is used to separate the classes, and the label ordering information is incorporated into learning the classifier by imputing the parallel hyperplanes with an order. Moreover, considering that a bag may contain instances not belonging to its class, for each bag, the instance which is nearest to the middle of the corresponding class is selected to learn the classifier. Compared with the existing single-instance OR work, MIOR is able to learn a more accurate OR classifier on multiple-instance data where only the bag label is available and the instance label is unknown. Extensive experiments show that MIOR outperforms the existing single-instance OR methods.

Haptic Collision Detection on Disjoint Objects with Overlapping and Inclusive Bounding Volumes.

This paper presents a method to alleviate performance degradation issues of Haptic Collision Detection when the Bounding Volumes or Bounding Volume Hierarchies of multiple disjoint objects are overlapping or inclusive and force the Haptic Collision Detection methods into narrow phase collision detection with all involved objects. The proposed method aims to generate tighter, mutually exclusive Bounding Volumes at the pre-processing stage, and to quickly cull irrelevant nearby objects at the broad phase to ensure that the Haptic Collision Detection methods will not be overloaded with unnecessary narrow phase collision detection. The proposed method is based on a hybrid representation of Bounding Volume and Space Partitioning and is implemented as an algorithm that automatically generates these new Bounding Volumes for disjoint objects, with details and corner cases discussed. A series of experiments based on real-life Haptic Collision Detection applications has been conducted. The results are analyzed and compared with those from an existing Haptic Collision Detection algorithm. The outcome demonstrates the capability of the proposed method in maintaining a stable Haptic Collision Detection performance under various challenging situations.

Attentive Observation Is Essential for the Misattribution of Agency to Self-Performance.

Recent studies have repeatedly demonstrated a false memory phenomenon in which people falsely remember having performed an action by oneself when in fact they have only observed the action by another person. We investigated the attentional effect to the action itself on the observation inflation. Fifty-four participants first performed and read actions (Phase 1); then, they observed the action video that showed another’s actions (Phase 2), some of which they had not performed in Phase 1. In the Phase 2, they were required to focus on either the actor’s performance (i.e., attentive observation condition) or irrelevant objects, which were presented in the background (i.e., inattentive observation condition) to modulate their attention. Around 2 weeks later, participants took a surprise source-memory test (Phase 3). In this phase, we asked them to judge whether they “performed,” “read,” or “not presented” the action in Phase 1. Three participants were removed from analysis, because they could not attend Phase 3 within 10–16 days after completion of the second phase. We found observation inflation only in the attentive condition, which contradicted the notions from other false memory studies that showed that attention to the target stimuli reduced false memory in general. We discussed the observation inflation mechanism from the perspective of the “like me” system, including the mirror neuron system, self-ownership, and self-agency.

Coordinating vision and action in natural behaviour: Differences in spatiotemporal coupling in everyday tasks.

Vision and action are tightly coupled in space and time: for many tasks we must look at the right place at the right time to gather the information that we need to complete our behavioural goals. Vision typically leads action by about 0.5 seconds in many natural tasks. However, the factors that influence this temporal coordination are not well understood, and variations have been found previously between two domestic tasks each with similar constraints: tea making and sandwich making. This study offers a systematic exploration of the factors that govern spatiotemporal coordination of vision and action within complex real-world activities. We found that the temporal coordination eye movements and action differed between tea making and sandwich making. Longer eye-hand latencies, more "look ahead" fixations and more looks to irrelevant objects were found when making tea than when making a sandwich. Contrary to previous suggestions, we found that the requirement to move around the environment did not influence the coordination of vision and action. We conclude that the dynamics of visual behaviour during motor acts are sensitive to the task and specific objects and actions required but not to the spatial demands requiring movement around an environment. (PsycINFO Database Record

Positive and Negative Consequences of Making Coffee among Breakfast Related Irrelevant Objects: Evidence from MCI, Dementia, and Healthy Ageing.

Previous studies have reported impairments in activities of daily living (ADL) performance in the presence of irrelevant but physically/functionally related objects in dementia patients. The aim of the present study was to increase our knowledge about the impact of the presence of contextually related non-target objects on ADL execution in patients with multi-domain mild cognitive impairment (MCI) and dementia. We compared ADL execution in patients with MCI, dementia, and healthy elderly participants under two experimental conditions: One in which the target objects were embedded with contextually related non-target items that constituted the object set necessary to complete two additional (but unrequired) ADL tasks related to the target task, and a second, control condition where target objects were surrounded by isolated objects (they never constituted a whole set needed to complete an alternative ADL task). Separate analysis of ADL errors associated with the target task versus errors involving the non-target objects revealed that, although the presence of contextually related objects facilitated the accomplishment of the target task, such a condition also led to errors involving the use of irrelevant objects in dementia and MCI. The presence of contextually related non-target items produces both positive and negative effects on ADL performance. These types of non-target objects might help to cue the retrieval of the action schema related to the target task, particularly in patients with MCI. In contrast, the presence of these objects might also lead to distraction in dementia and MCI. (JINS, 2017, 23, 481-492).

Robust unsupervised cluster matching for network data

Unsupervised cluster matching is a task to find matching between clusters of objects in different domains. Examples include matching word clusters in different languages without dictionaries or parallel sentences and matching user communities across different friendship networks. Existing methods assume that every object is assigned into a cluster. However, in real-world applications, some objects would not form clusters. These irrelevant objects deteriorate the cluster matching performance since mistakenly estimated matching affect on estimation of matching of other objects. In this paper, we propose a probabilistic model for robust unsupervised cluster matching that discovers relevance of objects and matching of object clusters, simultaneously, given multiple networks. The proposed method finds correspondence only for relevant objects, and keeps irrelevant objects unmatched, which enables us to improve the matching performance since the adverse impact of irrelevant objects is eliminated. With the proposed method, relevant objects in different networks are clustered into a shared set of clusters by assuming that different networks are generated from a common network probabilistic model, which is an extension of stochastic block models. Objects assigned into the same clusters are considered as matched. Edges for irrelevant objects are assumed to be generated from a noise distribution irrespective of cluster assignments. We present an efficient Bayesian inference procedure of the proposed model based on collapsed Gibbs sampling. In our experiments, we demonstrate the effectiveness of the proposed method using synthetic and real-world data sets, including multilingual corpora and movie ratings.