Experimental Evaluation of FLIR ATR Approaches—A Comparative Study

Li, B; Chellappa, Rama; Zheng, Qinfen; Der, Sandor Z.; Nasrabadi, Nasser M.; Chan, Lipchen Alex; Wang, L

doi:10.1006/cviu.2001.0938

Cited by 32 publications

(23 citation statements)

References 15 publications

(20 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Ref. [11] presented an empirical evaluation of several ATR algorithms for Forward-Looking-Infrared (FLIR) imagery such as convolutional neural network (CNN), principal component analysis (PCA), linear discriminant analysis (LDA), learning vector quantization (LVQ), modular neural networks (MNN) and two model-based algorithms, using Hausdorff metric-based matching and geometric hashing using a large database of real FLIR images. Ref.…”

Section: Introductionmentioning

confidence: 99%

Kernel sparse coding method for automatic target recognition in infrared imagery using covariance descriptor

Yang

Yao

Sun

et al. 2016

Infrared Physics & Technology

View full text Add to dashboard Cite

Section: Introductionmentioning

confidence: 99%

Kernel sparse coding method for automatic target recognition in infrared imagery using covariance descriptor

Yang

Yao

Sun

et al. 2016

Infrared Physics & Technology

View full text Add to dashboard Cite

“…IRST is also used in military applications. Thermal emission of gears operating on tanks or helicopters can, for example, be used to detect, track and lock-on to the target, so that many automatic target recognition (ATR) algorithms have been proposed [Li et al, 2001]. These segment and recognize vehicles, 45 ships and aircrafts [e.g.…”

mentioning

confidence: 99%

Analysis of Thermal Video for Coarse to Fine Particle Tracking in Volcanic Explosion Plumes

Bombrun¹,

Barra²,

Harris³

2015

Image Analysis

View full text Add to dashboard Cite

Abstract. This paper presents two algorithms of feature extraction and 10 segmentation. The first algorithm is applied to detect tens of thousands of targets moving at high velocities (100's m/s) and with different sizes, velocities, shapes and directions. Upon detection, we compute statistics for each of these parameters for each particle, without any assumption nor a priori information. The second algorithm was developed to detect a 15 slow moving convective cloud. The challenge was to follow the evolution of the contours of a heterogeneous element in front of a homogeneous but possibly moving background. These algorithms were applied on images acquired with thermal cameras with different settings (frame rate, frame size, focal length, instantaneous field of view). A case study concern-20 ing images of volcanic explosive events is finally presented. Volcanoes provide, during an eruption, a source of both ballistic ejecta and a convective plume of finer particles, gas and entrained air both of which can be imaged in the infrared wavelength. : Results ?

show abstract

“…Examples of boundary descriptions include distances of each boundary pixel from the centroid [26] and Fourier Descriptors [10]; examples of parts based approaches include decomposition into distinct surfaces [19], regional Principal Component Analysis (PCA) [20], [23], and 2D Discrete Cosine Transform (DCT) [10], [25]; examples of holistic descriptions include moment invariants [8], [10], [14], [31], and holistic PCA [16], [27].…”

Section: Introductionmentioning

confidence: 99%

“…Several studies [5], [16], [24] have compared the performance of various approaches for recognizing objects in IR images. However, all of these studies assumed good quality ROI localization or used a closed set identification setup; furthermore, in [16], [24] silhouette representations of objects was not used.…”

Section: Introductionmentioning

confidence: 99%

“…However, all of these studies assumed good quality ROI localization or used a closed set identification setup; furthermore, in [16], [24] silhouette representations of objects was not used. Hence the aim of this paper is to evaluate, on a common database, several combinations of feature extractors and classifiers for the purposes of silhouette verification, subject to imperfect ROI localization and other adverse conditions.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

On Classifying Silhouettes in Adverse Conditions

Sanderson

Gibbins

Proceedings of the 2004 Intelligent Sensors, Sensor Networks and Information Processing Conference, 2004.

View full text Add to dashboard Cite

We compare the performance of holistic and local feature approaches for the purpose of classifying silhouettes in adverse conditions (i.e. occlusions by other silhouettes, noise and imperfect localization by a Region of Interest algorithm, resulting in clipping and scale changes). Holistic feature extractors based on Hu's moment invariants and Principal Component Analysis (PCA) are coupled with a classifier based on gaussian densities, while a local feature extractor based on the 2D Hadamard Transform (HT) is coupled with a Gaussian Mixture Model (GMM) based classifier. Experiments show that the HT/GMM approach is relatively robust to clipping, scale changes and occlusions; however, in its current form it is highly sensitive to noise. The results further show that the moment based approach achieves relatively poor performance in advantageous conditions and is easily affected by clipping and occlusions; the PCA based approach is highly affected by scale changes and clipping, while being relatively robust to occlusions and noise.

show abstract

Experimental Evaluation of FLIR ATR Approaches—A Comparative Study

Cited by 32 publications

References 15 publications

Kernel sparse coding method for automatic target recognition in infrared imagery using covariance descriptor

Kernel sparse coding method for automatic target recognition in infrared imagery using covariance descriptor

Analysis of Thermal Video for Coarse to Fine Particle Tracking in Volcanic Explosion Plumes

On Classifying Silhouettes in Adverse Conditions

Contact Info

Product

Resources

About