Real-time classification of vehicles by type within infrared imagery

Kundegorski, Mikolaj E.; Akçay, Samet; Garanderie, Grégoire Payen de La; Breckon, Toby P.

doi:10.1117/12.2241106

Cited by 11 publications

(4 citation statements)

References 69 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Although larger datasets such as UCF-Crime [44] offer more scope, their non-uniform (internet-based) curation makes their use in the evaluation of anomaly detection for fixed camera surveillance deployment challenging. In addition, all datasets largely focus on visible-band (colour) imagery, with only very few containing both visible and infrared (thermal) imagery [9,17] despite the increasing prevalence of infrared (thermal) imagery within the operational surveillance context [6,13,21]. An additional challenge is the varying environmental conditions, which in itself affects both visible-band [24,29,30,44] and infrared-band anomaly detection alike [17].…”

Section: Introductionmentioning

confidence: 99%

Region-based Appearance and Flow Characteristics for Anomaly Detection in Infrared Surveillance Imagery

Gaus

Bhowmik

Isaac-Medina

et al. 2023

2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)

View full text Add to dashboard Cite

Anomaly detection is a classical problem within automated visual surveillance, namely the determination of the normal from the abnormal when operational data availability is highly biased towards one class (normal) due to both insufficient sample size, and inadequate distribution coverage for the other class (abnormal). In this work, we propose the dual use of both visual appearance and localized motion characteristics, derived from optic flow, applied on a per-region basis to facilitate object-wise anomaly detection within this context. Leveraging established object localization techniques from a region proposal network, optic flow is extracted from each object region and combined with appearance in the far infrared (thermal) band to give a 3-channel spatiotemporal tensor representation for each object (1 × thermal -spatial appearance; 2 × optic flow magnitude as x and y components -temporal motion). This formulation is used as the basis for training contemporary semi-supervised anomaly detection approaches in a region-based manner such that anomalous objects can be detected as a combination of appearance and/or motion within the scene. Evaluation is performed using the Long-Term infrared (thermal) Imaging (LTD) benchmark dataset against which successful detection of both anomalous object appearance and motion characteristics are demonstrated using a range of semi-supervised anomaly detection approaches.

show abstract

Section: Introductionmentioning

confidence: 99%

Region-based Appearance and Flow Characteristics for Anomaly Detection in Infrared Surveillance Imagery

Gaus

Bhowmik

Isaac-Medina

et al. 2023

2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)

View full text Add to dashboard Cite

show abstract

“…The use of both 2D visible-band and infrared-band camera imagery within many visual surveillance tasks are well established with numerous solutions spanning target detection, visual tracking and behaviour analytics. [1][2][3][4] More broadly, the complementary nature of using both colour and (far, long-wave) infrared has seen them extensively utilised in a range of computer vision applications for several decades. 5 Prior work on 2D visible-band and infrared-band imagery has spanned object tracking, 6,7 pedestrian detection, 8 stereo vision, 9 and autonomous platform deployment.…”

Section: Introductionmentioning

confidence: 99%

Visible to infrared transfer learning as a paradigm for accessible real-time object detection and classification in infrared imagery

Gaus

Bhowmik

Isaac-Medina

et al. 2020

Counterterrorism, Crime Fighting, Forensics, and Surveillance Technologies IV

Self Cite

View full text Add to dashboard Cite

show abstract

“…This expansion, both in demand and performance, has led to the broader consideration of computer vision applications in imagery domains beyond the visible spectrum, i.e. non-visible images such as infrared (thermal) [6], synthetic aperture radar (SAR) [7] and X-ray images [8]. Imaging within the non-visible spectrum provides sensing capabilities ranging from all-weather visibility, object temperature, material characteristics and sub-surface/object transparency.…”

Section: Introductionmentioning

confidence: 99%

Data Augmentation via Mixed Class Interpolation using Cycle-Consistent Generative Adversarial Networks Applied to Cross-Domain Imagery

Sasaki

Willcocks

Breckon

2020

Preprint

Self Cite

View full text Add to dashboard Cite

Machine learning driven object detection and classification within non-visible imagery has an important role in many fields such as night vision, all-weather surveillance and aviation security. However, such applications often suffer due to the limited quantity and variety of non-visible spectral domain imagery, where by contrast the high data availability in visible-band imagery readily enables contemporary deep learning driven detection and classification approaches. To address this problem, this paper proposes and evaluates a novel data augmentation approach that leverages the more readily available visible-band imagery via a generative domain transfer model. The model can synthesise large volumes of non-visible domain imagery by image translation from the visible image domain. Furthermore, we show that the generation of interpolated mixed class (non-visible domain) image examples via our novel Conditional CycleGAN Mixup Augmentation (C2GMA) methodology can lead to a significant improvement in the quality for non-visible domain classification tasks that otherwise suffer due to limited data availability. Focusing on classification within the Synthetic Aperture Radar (SAR) domain, our approach is evaluated on a variation of the Statoil/C-CORE Iceberg Classifier Challenge dataset and achieves 75.4% accuracy, demonstrating a significant improvement when compared against traditional data augmentation strategies.

show abstract

Real-time classification of vehicles by type within infrared imagery

Cited by 11 publications

References 69 publications

Region-based Appearance and Flow Characteristics for Anomaly Detection in Infrared Surveillance Imagery

Region-based Appearance and Flow Characteristics for Anomaly Detection in Infrared Surveillance Imagery

Visible to infrared transfer learning as a paradigm for accessible real-time object detection and classification in infrared imagery

Data Augmentation via Mixed Class Interpolation using Cycle-Consistent Generative Adversarial Networks Applied to Cross-Domain Imagery

Contact Info

Product

Resources

About