DETECT EXPLOSIVES BY INFRARED
IMAGERY USING MERGING ANOMALY
ALGORITHM AND IMAGE FEATURES

MS.F.FATHIMA

DETECT EXPLOSIVES BY INFRARED IMAGERY USING MERGING ANOMALY ALGORITHM AND IMAGE FEATURES

MS.F.FATHIMA
Assistant Professor, Dept. of Computer Science and Engineering, Bharath University, Chennai -600073

Related article at Pubmed, Scholar Google

Visit for more related articles at International Journal of Innovative Research in Computer and Communication Engineering

Abstract

A innovative tactic method used for Merging Different Anomaly Algorithm decisions with image space cell-structured features on a long wave infrared (LWIR) System with context of forward Looking (FL) buried explosive hazard detection along a road.A pre-screener is begin,widely helps to produce ensemble of trainable size contrast filters focusing in Universal Transverse Mercator(UTM) Space .Next,Features from different algorithm are mined from UTM confidence map along with average shift grouping in Universal Transverse Mercator (UTM) space. Features available in image chips demonstrating anomaly decisions from different algorithms are extracted from UTM confidence maps based on maximally stable extremal regions (MSERs) and Guassian mixture models(GMMs).Pre-scanner hits in UTM space are back projected into the video at multiple standoff distance and cell-Structured Local Binary Patterns(LBPs), Histogram of Gradients(HOGs) and Mean-variance descriptors are extracted. Experiments are conducted using Buried Volatiles with varying metal contents and depths in U.S.Army Test site. Results are extremely encouraging in FL imaging and show a significant decrease in the number of false alarms(Fas). Targests not currently detected by our system are also not detected by a manually under human visualization inspections

Keywords

Anomaly Detection,datafusion,explosive hazard detection,forward looking infrared imagery.

INTRODUCTION

Exposure in detection and Identification of explosive hazards in an exponentially important and these hazards are widely accountable to numerous deaths and wounds to both militants in the world. Arrangements which detect explosive hazards include Ground-Penetrating-Radar., Infrared (IR) Cameras. Research group investigated and develops a variety of algorithm for forward Looking (FL) vehicle mounted explosive detection technologies., FL Ground Penetrating Radar and Infrared imagery. Early approach organized independently to test detection at every place. In this paper to modify new image space descriptors at pre-scanner locations.This paper is focused on, discovering buried explosive materials in Forward Looking Perspective infrared imagery.Five important criteria themes exists in this work .at first, Improve to detect find targets in FL Perspective imagery for a moving platform. Later, Multiple different anomaly detection algorithms to find different image clusters into groups. Third ,Ensemble of detectors to operate anomaly decisions across frames. To identify the cell structured image features for FAR reduction .Finding targets e.g. buried explosives in forward looking approach.. This paper is widely structured around the flow of information. The image space anomaly detection algorithms and their individual temporal aggregations. The extraction of cell-structured image space features by the back projection of hits at standalone distance. It is achieved by UTM Algorithm decision maps. The information sources are then combined and receiver operating characteristic (ROC) curve analysis is used to determine Positive Detection(PD) and reduce False Alarm Rate(FAR) at different operating levels.

The main goal of this project is to detect and remediate the explosive hazards as these hazards are responsible for uncountable deaths and injuries to both civilians and soldiers throughout the world.This mission dedicated, but not partial to, discovering buried explosive materials in FL perspective infrared imagery. Five important general themes exist in this work.. First, it discusses a new way to detect difficult-to-find targets in FL viewpoint imagery for a moving platform (specifically ground-based).Multiple different anomaly detection algorithms are engaged to find different image artifacts at both the per-image level and across images (temporal change). Third, it discusses clustering within an image, across images and across an ensemble of detectors, as well as aggregation operators for combining anomaly decisions across frames.This project also presents a method of combining disparate system information that is multiple anomaly algorithm decisions and state-of-the-art cell-structured image-space features for FAR reduction. Lastly, the topic of finding targets (e.g., buried explosives) in LWIR is important in the remote sensing area.

To describe an algorithm based on an ensemble of trainable size-contrast filters and weighted mean shift clustering for the recognition of explosive materials (specifically buried targets). The core assumption made by this method is that evidence of buried targets appear in the LWIR imagery according to a known approximate size and that it exhibits high local image contrast. This approach is not restricted to LWIR .

METHODS PRE-SCREENER

Pre-screener hits in UTM space are back-projected into the video at multiple standoff distances and cellstructured local binary patterns (LBPs), histogram of gradients (HOGs) and mean-variance descriptors are extracted.

2.1 TRAINABLE SIZE CONTRAST FILTERS

Trainable size-contrast filters and weighted mean shift clustering for the recognition of explosive materials (specifically buried targets). The core assumption made by this method is that evidence of buried targets appears in the LWIR imagery according to a known approximate size and that it exhibits high local image contrast.

2.2 GAUSSIAN MIXTURE MODELS

To Detail an image space algorithm based on the concept of detecting local recent change (anomalies) in a road, specifically down track, while a vehicle is in motion. The method is based on Stauffer and Grimson’s Gaussian mixture model (GMM) algorithm. The GMM is a low-levelcomputer vision tool from the video surveillance domain for change detection.

2.3 MAXIMALLY STABLE EXTREMER REGION

The final image space anomaly detection algorithm explored does not assume a target shape. It like the other two algorithms discussed, is not restricted to LWIR. Themotivating idea behind the MSER approach is to search for buried explosive hazards in LWIR by looking for spatially coherent connected regions of expected approximate size using the maximally stable extremal regions (MSER) segmentation algorithm.

2.4 CELL STRUCTURED MEAN

The utility of two state-of-the-art texture features and a contrast feature is explored. Specifically, weighted mean shift cluster centers (pre-screener hits) are subject to a second round of classification based on image features. The reason for this second stage of classification is further reduction of the FAR.

CONCLUSION

This paper proposed a novel approach for combining multiple anomaly algorithm decisions with image space cell-structured features in a long wave infrared (LWIR) system in the context of forward looking (FL) buried explosive hazard detection along a road. FA reduction is the main benefit of the proposed work. It is predominately in the range of [0%, 85%] detection rates that this project shows the greatest improvement.

FUTURE WORK

In future work, it will investigate optimization strategies to learn system parameters currently specified by humans. While it has done some initial sensitivity analysis during development, not for all parameters, a more exhaustive and formal analysis is needed. Future work will also include the continued exploration of different image space algorithms, features and classifiers. Specifically, additional image space algorithms and perhaps other sensors are needed to address the current set of missed targets (what the pre-screener does not identity at any operating level).

Figures at a glance


Figure 1	Figure 2	Figure 3	Figure 4

References

Combination of Anomaly Algorithms and Image Features for Explosive Hazard Detection in Forward LookingInfrared Imagery,Derek T. Anderson,Kevin E. Stone,James M. Keller,ChristopherJ.Spain. ieee journal, applied earth observations and remote sensing,Feb 2012.

J. Farrell, T. Havens, K. C. Ho, J. M. Keller, T. Ton, D. Wong, and M. Soumekh, “Detection of explosive hazards using spectrum features from forward-looking ground-penetrating radar imagery,” presented at the Defense,Security, and Sensing Conf., Orlando, FL, 2011.

H. Frigui and P. Gader, “Detection and discrimination of land mines in ground-penetrating radar based on edgehistogram descriptors and a possibilistic-nearest neighbor classifier,” IEEE Trans. Fuzzy Syst., vol. 17, no. 1, pp.185–199, Feb. 2009.

T. C. Havens, J.M. Keller, K. C. Ho, T. T. Ton, D. C. Wong, and M. Soumekh, “Narrow-band processing andfusion approach for explosive hazard detection in FLGPR,” presented at the Defense, Security, and Sensing Conf.,Orlando, FL, 2011.

T. C. Havens, K. C. Ho, J. M. Keller, M. Popescu, T. T. Ton, D. C. Wong, and M. Soumekh, “Locally adaptivedetection algorithm for forward-looking ground-penetrating radar,” in Detection and Sensing of Mines, ExplosiveObjects, and Obscured Targets XV, Orlando, 2010.

T. C. Havens,C. J.Spain,K.C.Ho, J.M. Keller,T.T.Ton,D.C.Wong, and M. Soumekh, “Improved detection andfalse alarmrejection using FLGPR and color imagery in a forward-looking system,” in Detection and Sensing ofMines, Explosive Objects, and Obscured Targets [6] C. R. Ratto, P. A. Torrione, and L.M. Collins, “Exploitingground-penetrating radar phenomenology in a context-dependent framework for landmine detection anddiscrimination,” IEEE Trans. Geosci. Remote Sens., vol. 49, no. 5, pp. 1689–1700, May 2011.

F. Soldovieri, O. Lopera, and S. Lambot, “Combination of advanced inversion techniques for an accurate targetlocalization via GPR for demining applications,” IEEE Trans. Geosci. Remote Sens., vol. 49, no. 1, pp. 451–461,Jan. 2011.

T. Wang, J. M. Keller, P. D. Gader, and O. Sjahputera, “Frequency subband processing and feature analysis offorward-looking groundpenetrating radar signals for land-mine detection,” IEEE Trans. Geosci. Remote Sens., vol.45, no. 3, pp. 718–729, 2007.

T. Wang, O. Sjahputera, J. M. Keller, and P. D. Gader, “Feature analysis for forward-looking landmine detectionusing GPR,” in Detection and Remediation Technologies for Mines and Minelike Targets X, Orlando, FL, 2005,vol. 5794, pp. 1233–1244.

T.Wang, O. Sjahputera, J. M. Keller, and P. D. Gader, “Landmine detection using forward-looking GPR withobject tracking,” in Detection and Remediation Technologies for Mines and Minelike Targets X, Orlando, FL, 2005,vol. 5794, pp. 1080–1088.

K. C. Ho, L. Carin, P. D. Gader, and J. N.Wilson, “An investigation of using the spectral characteristics fromground penetrating radar for landmine/clutter discrimination,” IEEE Trans. Geosci. Remote Sens., vol. 46, no. 4, pp.1177–1191, Apr. 2008.

D. T. Anderson, J. M. Keller, and O. Sjahputera, “Algorithm fusion in forward-looking longwave infrared imagery for buried explosive hazard detection,” presented at the Defense, Security, and Sensing Conf., Orlando, FL,2011.

M. Popescu, K. E. Stone, and J. M. Keller, “Detection of targets in forward-looking infrared imaging using amultiple instance learning framework,” presented at the Defense, Security, and Sensing Conf., Orlando, FL, 2011.

M. Popescu, K. Stone, T. Havens, D. Ho, and J. Keller, “Anomaly detection in forward looking infrared imaging using one-class classifiers,” in Detection and Sensing of Mines, Explosive Objects, and Obscured TargetsXV, Orlando, FL, 2010, pp. 76642B–76642B-8.

C. J. Spain, D. T. Anderson, J. M. Keller, M. Popescu, and K. E. Stone, “Gaussian mixture models for measuring local change down-track in LWIR imagery for explosive hazard detection,” presented at the Defense,Security, and Sensing Conf., Orlando, FL, 2011.

K. Stone, J. M. Keller, M. Popescu, T. C. Havens, and K. C. Ho, “Forward looking anomaly detection via fusion of infrared and color imagery,” in Detection and Sensing of Mines, Explosive Objects, and Obscured TargetsXV, Orlando, FL, 2010, pp. 766425–766425-12.

K. E. Stone, J. M. Keller, M. Popescu, and C. J. Spain, “Buried explosive hazard detection using forwardlookinglong-wave infrared imagery,” presented at the Defense, Security, and Sensing Conf., Orlando, FL, 2011.

N. T. Thanh, H. Sahli, and D. N. Hao, “Infrared thermography for buried landmine detection: Inverse roblemsetting,” IEEE Trans. Geosci. Remote Sens., Dec. 2008.

N. Xiang and J. M. Sabatier, “Laser dopplervibrometer-based acoustic landmine detection using the fast Msequencetransform,” IEEE Geosci. Remote Sens. Lett., vol. 1, no. 4, Oct. 2004.

J. M. Sabatier and N. Xiang, “An investigation of acoustic-to-seismic coupling to detect buriedantitank landmines,” IEEE Trans. Geosci. Remote Sens., vol. 39, no. 6, pp. 1146–1154, Jun. 2001.