Visual Detection Task Research Articles

Hierarchical reasoning based on perception action cycle for visual question answering

Recent visual question answering (VQA) frameworks employ different attention modules to derive a correct answer. The concept of attention is heavily established in human cognition, which led to its magnificent success in deep neural networks. In this study, we aim to consider a VQA framework that utilizes human biological and psychological concepts to achieve a good understanding of vision and language modalities. In this view, we introduce a hierarchical reasoning method based on the perception action cycle (HIPA) framework to tackle VQA tasks. The perception action cycle (PAC) explains how humans learn about and interact with their surrounding world. The proposed framework integrates the reasoning process of multi-modalities with the concepts introduced in PAC in multiple phases. It comprehends the visual modality through three phases of reasoning: object-level attention, organization, and interpretation. In addition, it comprehends the language modality through word-level attention, interpretation, and conditioning. Subsequently, vision and language modalities are interpreted dependently in a cyclic and hierarchical way throughout the entire framework. For further assessment of the generated visual and language features, we argue that image–question pairs of the same answer ought to eventually have similar visual and language features. As a result, we conduct visual and language feature evaluation experiments using metrics such as the standard deviation of cosine similarity and of Manhattan distance. We show that employing PAC in our framework improves the standard deviation compared with other VQA frameworks. For further assessment, we also test the novel proposed HIPA on the visual relationship detection (VRD) task. The proposed method achieves state-of-the-art results on the TDIUC and VRD datasets and obtains competitive results on the VQA 2.0 dataset. The code is available: github.com/Safaa1113/HiPA-Framework.

Neural processing of bottom-up perception of biological motion under attentional load

Considering its importance for one’s survival and social significance, biological motion (BM) perception is assumed to occur automatically. Previous behavioral results showed that task-irrelevant BM in the periphery interfered with task performance at the fovea. Under selective attention, BM perception is supported by a network of regions including the occipito-temporal (OTC), parietal, and premotor cortices. Retinotopy studies that use BM stimulus showed distinct maps for its processing under and away from selective attention. Based on these findings, we investigated how bottom-up perception of BM would be processed in the human brain under attentional load when it was shown away from the focus of attention as a task-irrelevant stimulus. Participants (N = 31) underwent an fMRI study in which they performed an attentionally demanding visual detection task at the fovea while intact or scrambled point light displays of BM were shown at the periphery. Our results showed the main effect of attentional load in fronto-parietal regions and both univariate activity maps and multivariate pattern analysis results support the attentional load modulation on the task-irrelevant peripheral stimuli. However, this effect was not specific to intact BM stimuli and was generalized to motion stimuli as evidenced by the motion-sensitive OTC involvement during the presence of dynamic stimuli in the periphery. These results confirm and extend previous work by showing that task-irrelevant distractors can be processed by stimulus-specific regions when there are enough attentional resources available. We discussed the implications of these results for future studies.

Optogenetic activation of visual thalamus generates artificial visual percepts.

The lateral geniculate nucleus (LGN), a retinotopic relay center where visual inputs from the retina are processed and relayed to the visual cortex, has been proposed as a potential target for artificial vision. At present, it is unknown whether optogenetic LGN stimulation is sufficient to elicit behaviorally relevant percepts, and the properties of LGN neural responses relevant for artificial vision have not been thoroughly characterized. Here, we demonstrate that tree shrews pretrained on a visual detection task can detect optogenetic LGN activation using an AAV2-CamKIIα-ChR2 construct and readily generalize from visual to optogenetic detection. Simultaneous recordings of LGN spiking activity and primary visual cortex (V1) local field potentials (LFP) during optogenetic LGN stimulation show that LGN neurons reliably follow optogenetic stimulation at frequencies up to 60 Hz, and uncovered a striking phase locking between the V1 local field potential (LFP) and the evoked spiking activity in LGN. These phase relationships were maintained over a broad range of LGN stimulation frequencies, up to 80 Hz, with spike field coherence values favoring higher frequencies, indicating the ability to relay temporally precise information to V1 using light activation of the LGN. Finally, V1 LFP responses showed sensitivity values to LGN optogenetic activation that were similar to the animal's behavioral performance. Taken together, our findings confirm the LGN as a potential target for visual prosthetics in a highly visual mammal closely related to primates.

Age-related hearing loss associated with differences in the neural correlates of feature binding in visual working memory

The underlying neural mechanisms underpinning the association between age-related hearing loss (ARHL) and dementia remain unclear. A limitation has been the lack of functional neuroimaging studies in ARHL cohorts to help clarify this relationship. In the present study, we investigated the neural correlates of feature binding in visual working memory with ARHL (controls = 14, mild HL = 21, and moderate or greater HL = 23). Participants completed a visual change detection task assessing feature binding while their neural activity was synchronously recorded via high-density electroencephalography. There was no difference in accuracy scores for ARHL groups compared to controls. There was increased electrophysiological activity in those with ARHL, particularly in components indexing the earlier stages of visual cognitive processing. This activity was more pronounced with more severe ARHL and was associated with maintained feature binding. Source space (sLORETA) analyses indicated greater activity in networks modulated by frontoparietal and temporal regions. Our results demonstrate there may be increased involvement of neurocognitive control networks to maintain lower-order neurocognitive processing disrupted by ARHL.

The role of middle frontal gyrus in working memory retrieval by the effect of target detection tasks: a simultaneous EEG-fMRI study.

Maintained working memory (WM) representations have been shown to influence visual target detection selection, while the effect of the visual target detection process on WM retrieval remains largely unknown. In the current research, we used the dual-paradigm of the visual target detection task and the delayed matching task (DMT), which contained the following four conditions: the match condition: the DMT target contained the detection target; the mismatch condition: the DMT target contained the detection distractor; the neutral condition: only the detection target was presented; the catch condition: only the DMT target was presented. Twenty-six subjects were recruited in the experiment with simultaneous EEG-fMRI data. Behaviorally, faster responses were found in the mismatch condition than in the match and neutral conditions. The EEG data found a greater parieto-occipital N1 component in the mismatch condition compared to the neutral condition, and a greater frontal N2 component in the match condition than in the mismatch condition. Moreover, compared to the match and neutral conditions, weaker activations of the bilateral middle frontal gyrus (MFG) were observed in the mismatch condition. And the representational similarity analysis (RSA) revealed significant differences in the representational patterns of the bilateral MFG between mismatch and match conditions, as well as in the representational patterns of the left MFG between mismatch and neutral conditions. Additionally, the left MFG may be the brain source of the N1 component in the mismatch condition. These findings suggest that the mismatch between the DMT target and detection target affects early attention allocation and attentional control in WM retrieval, and the MFG may play an important role in WM retrieval by the effect of the target detection task. In conclusion, our work deepens the understanding of the neural mechanisms by which visual target detection affects WM retrieval.

Contrast detection is enhanced by deterministic, high-frequency transcranial alternating current stimulation with triangle and sine waveform.

Stochastic resonance (SR) describes a phenomenon where an additive noise (stochastic carrier-wave) enhances the signal transmission in a nonlinear system. In the nervous system, nonlinear properties are present from the level of single ion channels all the way to perception and appear to support the emergence of SR. For example, SR has been repeatedly demonstrated for visual detection tasks, also by adding noise directly to cortical areas via transcranial random noise stimulation (tRNS). When dealing with nonlinear physical systems, it has been suggested that resonance can be induced not only by adding stochastic signals (i.e., noise) but also by adding a large class of signals that are not stochastic in nature that cause "deterministic amplitude resonance" (DAR). Here, we mathematically show that high-frequency, deterministic, periodic signals can yield resonance-like effects with linear transfer and infinite signal-to-noise ratio at the output. We tested this prediction empirically and investigated whether nonrandom, high-frequency, transcranial alternating current stimulation (tACS) applied to the visual cortex could induce resonance-like effects and enhance the performance of a visual detection task. We demonstrated in 28 participants that applying 80-Hz triangular-waves or sine-waves with tACS reduced the visual contrast detection threshold for optimal brain stimulation intensities. The influence of tACS on contrast sensitivity was equally effective to tRNS-induced modulation, demonstrating that both tACS and tRNS can reduce contrast detection thresholds. Our findings suggest that a resonance-like mechanism can also emerge when deterministic electrical waveforms are applied via tACS.NEW & NOTEWORTHY Our findings extend our understanding of neuromodulation induced by noninvasive electrical stimulation. We provide the first evidence showing acute online benefits of transcranial alternating current stimulation (tACS)triangle and tACSsine targeting the primary visual cortex (V1) on visual contrast detection in accordance with the resonance-like phenomenon. The "deterministic" tACS and "stochastic" high-frequency-transcranial random noise stimulation (tRNS) are equally effective in enhancing visual contrast detection.

Acute exercise has specific effects on the formation process and pathway of visual perception in healthy young men.

Visual perception is formed over time through the formation process and visual pathway. Exercise improves visual perception, but it is unclear whether exercise modulates nonspecifically or specifically the formation process and pathway of visual perception. Healthy young men performed the visual detection task in a backward masking paradigm before and during cycling exercise at a mild intensity or rest (control). The task presented gratings of a circular patch (target) and annulus (mask) arranged concentrically as a visual stimulus and asked if the presence and striped pattern (feature) of the target were detected. The relationship between the orientations of the gratings of the target and the mask included iso-orientation and orthogonal orientation to investigate the orientation selectivity of the masking effect. The masking effect was evaluated by perceptual suppressive index (PSI). Exercise improved feature detection (∆PSI; Exercise: -20.6%, Control: 1.7%) but not presence detection (∆PSI; Exercise: 8.9%, Control: 29.6%) compared to the control condition, and the improving effect resulted from the attenuation of the non-orientation-selective (∆PSI; Exercise: -29.0%, Control: 16.8%) but not orientation-selective masking effect (∆PSI; Exercise: -3.1%, Control: 11.7%). These results suggest that exercise affects the formation process of the perceptual feature of the target stimulus by suppressively modulating the neural networks responsible for the non-orientation-selective surround interaction in the subcortical visual pathways, whose effects are inherited by the cortical visual pathways necessary for perceptual image formation. In conclusion, our findings suggest that acute exercise improves visual perception transiently through the modulation of a specific formation process of visual processing.

Dual attentional transformer for video visual relation prediction

Video visual relation detecti on (VidVRD) is to detect visual relations among instances as well as the trajectories of the corresponding subjects and objects in the video. Most current works improve the accuracy of tracking the objects but neglect the other key challenge, predicting the reliable visual relations in the videos, a vital meant for downstream tasks further. In this paper, we propose a dual attentional transformer network (VRD-DAT) for predicting the visual relations, also known as the predicates, in multi-relation videos. Specifically, our network first respectively targets modeling action visual predicates (Act-T) and spatial locating visual relations (Spa-T) via two parallel visual transformer structures simultaneously. Then, an attentional weighting module obtains the final precise merged visual relations. We conduct extensive experiments on two public datasets, ImageNet-VidVRD and VidOR, to demonstrate our model is capable of outperforming other state-of-the-art methods effectively on the task of video visual relation prediction. Quantitative and qualitative results also show that with more accurate visual relations, the performance of the video visual relation detection task can be further boosted.

Topological insights into the neural basis of flexible behavior

It is widely accepted that there is an inextricable link between neural computations, biological mechanisms, and behavior, but it is challenging to simultaneously relate all three. Here, we show that topological data analysis (TDA) provides an important bridge between these approaches to studying how brains mediate behavior. We demonstrate that cognitive processes change the topological description of the shared activity of populations of visual neurons. These topological changes constrain and distinguish between competing mechanistic models, are connected to subjects' performance on a visual change detection task, and, via a link with network control theory, reveal a tradeoff between improving sensitivity to subtle visual stimulus changes and increasing the chance that the subject will stray off task. These connections provide a blueprint for using TDA to uncover the biological and computational mechanisms by which cognition affects behavior in health and disease.

Emotional distraction and facilitation across sense and time.

Emotional stimuli can disrupt or enhance task performance according to factors that are presently poorly understood. One potentially important determinant is the sensory modality involved. In unimodal visual paradigms (visual task-irrelevant stimuli during a visual task) emotional stimuli frequently produce distraction effects; however, the effects across modalities appear more complex and may also depend on factors related to stimulus timing. It is entirely unclear how task-irrelevant visual stimuli impact auditory task performance in cross-modal paradigms. This project explored task performance as a function of sensory modality, emotional valence, and stimulus timing. In Study 1, participants (N = 50) completed a visual stimulus detection task in the presence of task-irrelevant negative and neutral images and sounds. Accuracy was disrupted in the presence of visual but not auditory emotional stimuli, particularly when the target and task-irrelevant stimulus appeared simultaneously. In Study 2, participants (N = 38) completed an equivalent auditory stimulus detection task. In sharp contrast to the effects observed with visual targets, response times and accuracy were enhanced in the presence of auditory emotional stimuli at the first timepoint but disrupted at later timepoints. However, there was no effect of task-irrelevant visual stimuli on auditory task performance. These findings demonstrate the importance of both sensory modality and timing in determining how emotional stimuli affect task performance and lay the groundwork for future studies examining the interaction between emotional and attentional processes. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Tool use moves the peri-personal space from the hand to the tip of the tool.

In this study, we used a visual target detection task to investigate three hypotheses about how the peri-personal space is extended after tool-use training: Addition, Extension, and Projection hypotheses. We compared the target detection performance before and after tool-use training. In both conditions, the participants held a hockey stick-like tool in their hands during the detection task. Furthermore, we added the no-tool-holding condition to the experimental design. In the no-tool-holding condition, a peri-hand space advantage in the visual target detection task was observed. When the participants held the tool with their hands, this peri-hand space advantage was lost. Furthermore, there was no peri-tool space advantage before tool training. After tool training, the peri-tool space advantage was observed. However, after tool training, the advantage of the peri-hand space was not observed. This result suggested that the peri-hand advantage was reduced by simply holding the tool because the participants lost the functionality of their hands. Furthermore, tool-use training improved detection performance only in the peri-tool space. Thus, these results supported the projection hypothesis that the peri-personal space advantage would move from the body to the functional part of the tool.

Deviance Distraction and Stimulus-Specific Adaptation in the Somatosensory Cortex Reduce with Experience.

Automatic detection of a surprising change in the sensory input is a central element of exogenous attentional control. Stimulus-specific adaptation (SSA) is a potential neuronal mechanism detecting such changes and has been robustly described across sensory modalities and different instances of the ascending sensory pathways. However, little is known about the relationship of SSA to perception. To assess how deviating stimuli influence target signal detection, we used a behavioral cross-modal paradigm in mice and combined it with extracellular recordings from the primary somatosensory whisker cortex. In this paradigm, male mice performed a visual detection task while task-irrelevant whisker stimuli were either presented as repetitive "standard" or as rare deviant stimuli. We found a deviance distraction effect on the animals' performance: Faster reaction times but worsened target detection was observed in the presence of a deviant stimulus. Multiunit activity and local field potentials exhibited enhanced neuronal responses to deviant compared with standard whisker stimuli across all cortical layers, as a result of SSA. The deviant-triggered behavioral distraction correlated with these enhanced neuronal deviant responses only in the deeper cortical layers. However, the layer-specific effect of SSA on perception reduced with increasing task experience as a result of statistical distractor learning. These results demonstrate a layer-specific involvement of SSA on perception that is susceptible to modulation over time.SIGNIFICANCE STATEMENT Detecting sudden changes in our immediate environment is behaviorally relevant and important for efficient perceptual processing. However, the connection between the underpinnings of cortical deviance detection and perception remains unknown. Here, we investigate how the cortical representation of deviant whisker stimuli impacts visual target detection by recording local field potential and multiunit activity in the primary somatosensory cortex of mice engaged in a cross-modal visual detection task. We find that deviant whisker stimuli distract animals in their task performance, which correlates with enhanced neuronal responses for deviants in a layer-specific manner. Interestingly, this effect reduces with the increased experience of the animal as a result of distractor learning on statistical regularities.

Nonlinear dynamics of EEG responses to unmanned vehicle visual detection with different levels of task difficulty

The main objective of this study was to examine the presence of chaos in the EEG recordings of brain activity under simulated unmanned ground vehicle visual detection scenarios with different levels of task difficulty. One hundred and fifty people participated in the experiment and completed four visual detection task scenarios: (1) change detection, (2) a threat detection task, (3) a dual-task with different change detection task rates, and (4) a dual-task with different threat detection task rates. We used the largest Lyapunov exponent and correlation dimension of the EEG data and performed 0–1 tests on the EEG data. The results revealed a change in the level of nonlinearity in the EEG data corresponding to different levels of cognitive task difficulty. The differences in EEG nonlinearity measures among the studied levels of task difficulty, as well as between a single task scenario and a dual-task scenario, have also been assessed. The results increase our understanding of the nature of unmanned systems' operational requirements.

Visual perceptual load and processing of somatosensory stimuli in primary and secondary somatosensory cortices

Load theory assumes that neural activation to distractors in early sensory cortices is modulated by the perceptual load of a main task, regardless of whether task and distractor share the same sensory modality or not. While several studies have investigated the question of load effects on distractor processing in early sensory areas, there is no functional magnetic resonance imaging (fMRI) study regarding load effects on somatosensory stimuli. Here, we used fMRI to investigate effects of visual perceptual load on neural responses to somatosensory stimuli applied to the wrist in a study with 44 participants. Perceptual load was manipulated by an established sustained visual detection task, which avoided simultaneous target and distractor presentations. Load was operationalized by detection difficulty of subtle or clear color changes of one of 12 rotating dots. While all somatosensory stimuli led to activation in somatosensory areas SI and SII, we found no statistically significant difference in brain activation to these stimuli under high compared to low sustained visual load. Moreover, exploratory Bayesian analyses supported the absence of differences. Thus, our findings suggest a resistance of somatosensory processing to at least some forms of visual perceptual load, possibly due to behavioural relevance of discrete somatosensory stimuli and separable attentional resources for the somatosensory and visual modality.

Fluoxetine degrades luminance perceptual thresholds while enhancing motivation and reward sensitivity.

Selective serotonin reuptake inhibitors (SSRIs) increase serotonin activity in the brain. While they are mostly known for their antidepressant properties, they have been shown to improve visual functions in amblyopia and impact cognitive functions ranging from attention to motivation and sensitivity to reward. Yet, a clear understanding of the specific action of serotonin to each of bottom-up sensory and top-down cognitive control components and their interaction is still missing. To address this question, we characterize, in two adult male macaques, the behavioral effects of fluoxetine, a specific SSRI, on visual perception under varying bottom-up (luminosity, distractors) and top-down (uncertainty, reward biases) constraints while they are performing three different visual tasks. We first manipulate target luminosity in a visual detection task, and we show that fluoxetine degrades luminance perceptual thresholds. We then use a target detection task in the presence of spatial distractors, and we show that under fluoxetine, monkeys display both more liberal responses as well as a degraded perceptual spatial resolution. In a last target selection task, involving free choice in the presence of reward biases, we show that monkeys display an increased sensitivity to reward outcome under fluoxetine. In addition, we report that monkeys produce, under fluoxetine, more trials and less aborts, increased pupil size, shorter blink durations, as well as task-dependent changes in reaction times. Overall, while low level vision appears to be degraded by fluoxetine, performances in the visual tasks are maintained under fluoxetine due to enhanced top-down control based on task outcome and reward maximization.

Pre-stimulus gamma power in human posteromedial cortex shows supra-modal mechanisms in predicting the amplitude and latency of task-induced suppression.

Upon repetitively performing the same well-practiced task on identical bottom-up stimuli, our performance still varies. Although it has been well documented that elevated pre-stimulus baseline activity in the human default-mode network impairs the subsequent task performance, it remains unknown (i) the fine-grained temporal dynamics and (ii) whether the underlying neural dynamics are supra-modal or modality-specific. We utilized intracranial recordings in the human posteromedial cortex (PMC) during a simple visual and an auditory detection task. Our findings suggested that the pre-stimulus gamma power in PMC predicted the subsequent task performance. Critically, the higher the pre-stimulus gamma power, the longer it took for it to be suppressed, and the less suppressed it was during the task performance, which eventually resulted in deleterious effects on task performance, i.e. longer reaction times. These fine-grained temporal dynamics were consistent between the visual and auditory simple detection task. In addition, a direct comparison between the visual and auditory modality showed that the between-modality difference emerged during the recovery period from the maximal gamma suppression back to the baseline. Taken together, the present results contribute novel spatio-temporal mechanisms in human PMC on how simple detection performance varies across multiple repetitions, irrespective of the sensory modality involved.

Boosting Relationship Detection in Images with Multi-Granular Self-Supervised Learning

Visual and spatial relationship detection in images has been a fast-developing research topic in the multimedia field, which learns to recognize the semantic/spatial interactions between objects in an image, aiming to compose a structured semantic understanding of the scene. Most of the existing techniques directly encapsulate the holistic image feature plus the semantic and spatial features of the given two objects for predicting the relationship, but leave the inherent supervision derived from such structured and thorough image understanding under-exploited. Specifically, the inherent supervision among objects or relations within an image can span different granularities in this hierarchy including, from simple to comprehensive, (1) the object-based supervision that captures the interaction between the semantic and spatial features of each individual object, (2) the inter-object supervision that characterizes the dependency within the relationship triplet ( <subject-predicate-object> ), and (3) the inter-relation supervision that exploits contextual information among all relationship triplets in an image. These inherent multi-granular supervisions offer a fertile ground for building self-supervised proxy tasks. In this article, we compose a trilogy of exploring the multi-granular supervision in the sequence from object-based, inter-object, and inter-relation perspectives. We integrate the standard relationship detection objective with a series of proposed self-supervised proxy tasks, which is named as Multi-Granular Self-Supervised learning (MGS). Our MGS is appealing in view that it is pluggable to any neural relationship detection models by simply including the proxy tasks during training, without increasing the computational cost at inference. Through extensive experiments conducted on the SpatialSense and VRD datasets, we demonstrate the superiority of MGS for both spatial and visual relationship detection tasks.

The role of colour in the visual perception of trains and implications for railway worker safety

In the rail environment, Lookouts protect workers on track by visually monitoring the environment and detecting approaching trains. It is therefore critically important that trains are easily detected by Lookouts. The specificity of standards regarding front-of-train colour designs are varied and ambiguous. As well, limited guidance is found in the body of literature on conspicuity regarding how it applies to the rail and broader transport industry. In visual perception research, conflicting views exist on whether colour is important in determining object conspicuity. The present study seeks to address this research gap by employing a novel use of virtual reality (VR) technology, to simulate a train detection task for Lookouts. The experiments investigated (1) whether train conspicuity differs for varying train types under different ambient conditions, and (2) whether colour is important for improving train conspicuity. We found differences in the time taken to detect different train types, providing evidence that train conspicuity is affected by front-of-train design. However, the observed conspicuity differences between train types were not due to chromatic colour differences. The study findings raise interesting questions regarding the role of colour in visual detection tasks and demonstrates a novel application of empirical research to address rail industry issues.

Relationship between Reaction Times and Post-COVID-19 Symptoms Assessed by a Web-Based Visual Detection Task.

Long-COVID is a clinical condition in which patients affected by SARS-CoV-2 usually report a wide range of physical and cognitive symptoms from 3 to 6 months after the infection recovery. The aim of the current study was to assess the link between self-reported long-COVID symptoms and reaction times (RTs) in a self-administered Visual Detection Task (VDT) in order to identify the predictor symptoms of the slowing in reaction times to determine attention impairment. In total, 362 participants (age (mean ± S.D.: 38.56 ± 13.14); sex (female-male: 73.76-26.24%)) responded to a web-based self-report questionnaire consisting of four sections: demographics, disease-related characteristics, and medical history questions. The final section consisted of a 23 item 5-point Likert-scale questionnaire related to long-term COVID-19 symptoms. After completing the questionnaire, subjects performed a VDT on a tablet screen to assess reaction times (RTs). An exploratory factorial analysis (EFA) was performed on the 23 long-COVID symptom questions, identifying 4 factors (cognition, behavior, physical condition, presence of anosmia and/or ageusia). The most important predictors of RTs were cognition and physical factors. By dissecting the cognitive and physical factors, learning, visual impairment, and headache were the top predictors of subjects' performance in the VDT. Long-COVID subjects showed higher RTs in the VDT after a considerable time post-disease, suggesting the presence of an attention deficit disorder. Attention impairment due to COVID-19 can be due to the presence of headaches, visual impairments, and the presence of cognitive problems related to the difficulty in learning new activities. The link between the slowing of reaction times and physical and cognitive symptoms post-COVID-19 suggests that attention deficit disorder is caused by a complex interaction between physical and cognitive symptoms. In addition, the study provides evidence that RTs in a VDT represent a reliable measure to detect the presence of long-COVID neurological sequelae.

Rethinking unsupervised texture defect detection using PCA

In recent times, deep learning frameworks have become the preferred solution for visual object detection tasks, including texture defect detection. However, in real-world industrial applications, such as operation and maintenance of gas-insulated switchgear (GIS) equipment, collecting a large number of samples in advance is difficult because opening GIS equipment affects its normal operation; moreover, only four real-world GIS equipment defect samples are available currently. Most deep learning methods perform poorly when trained using a small number of labelled samples or when there is a lack of computational resources. In addition, the features extracted by multi-layer convolution and down-sampling are often unsuitable for small-size and low-contrast defects. However, the structure of the repetitive textured surface is simple and easy to model. Hence, this paper proposes a texture-defect detection method using principal components analysis (PCA) and histogram-based outlier score (HBOS) that requires only a small number of unlabelled samples and low computational complexity. First, PCA is utilized to extract features from the training patches to enhance the difference between the defect and background regions. Then, non-saliency suppression is applied for feature fusion to obtain the saliency map. Finally, the saliency value is converted to the defect score, and the defect is located by HBOS. Experiments on two public datasets and a GIS simulation equipment demonstrate that the proposed method is suitable for product surface defect detection and operation and maintenance in power distribution equipment. Moreover, compared with traditional texture feature-based and deep learning methods, the proposed method yields better results for small-size and low-contrast defects with only four unlabelled training samples per class.