Image Fusion of the Terahertz-Visual NAECON Grand Challenge Data Erik Blasch Air Force Research Lab Rome, NY, 13441 erik.blasch@rl.af.mil Zheng Liu Toyota Technological Institute Nagoya, Japan 468-8511 zhengliu@toyota-ti.ac.jp Doug Petkie Wright State Univ. Dayton, OH 45324 doug.petkie@wright.edu Robert Ewing AFRL Sensors Directorate Dayton, OH 45433 robert.ewing@wpafb.af.mil Gernot Pomrenke AFRL/AFOSR Arlington, VA 22203 gernot.pomrenke@afosr.af.mil Kitt Reinhardt AFRL/AFOSR Arlington, VA 22203 kitt.reinhardt@afosr.af.mil Abstract—Terahertz (THz) sensing has been developed over the past three decades for concealed weapons detection, medical imaging, and non-destructive evaluation; however methods for THz image exploitation have not been well reported. We test a multiscale image fusion algorithm for the 2011 IEEE National Aerospace and Electronics Conf. (NAECON) Grand Challenge which consists of Terahertz (THz) and visual images. The study consists of image characterization (signals distribution), image processing (data fusion), and image analysis (edge detection). We found that THz image characterization did not necessarily follow a distinct Gaussian distribution, THz imagery fusion with visual data supported target detection, and that image analysis enhanced target assessment. For the initial experiment, we assess the target segmentation through edge detection, image fusion results, and image fusion quality assessment. The preliminary image exploitation and fusion results can further develop THz collection over clothing-obscured concealed weapons imaging, parameter optimization, and targeting evaluation. Keywords: Image fusion, THz, Visual, Metrics, Evaluation, Concealed weapons detection I. INTRODUCTION There has been significant interest in the use of the submillimeter (SMM) or the terahertz (THz) spectral region for imaging over 30 years [1, 2, 3]. This interest has been driven by this spectral region’s unique combination of spatial resolution and penetration through many dielectric materials. One goal of these efforts has been to develop standoff imaging systems that can produce optical-like images for concealed weapon/object detection on personnel. [4, 5] The detection of concealed weapons and explosives represents one of the most daunting problems facing the military and civilian law enforcement personnel and new methods are needed to exploit THz imagery for object identification. The consideration of phenomenology in the THz is particularly interesting and important because the THz occupies an intermediate region between the optical and the microwave, shown in Fig. 1. The large majority of materials (excluding mirrors and mirror-like surfaces) are rough on an optical scale and smooth on typical microwave radar scales. At general THz wavelengths and 640 GHz in particular, there is a transition region between the specular reflections, or glint, that are associated with smooth surfaces and the diffuse reflections with rough surfaces [6]. As a result, complex images, which are hard to predict because they depend upon target texture details, are often observed. Moreover, because THz is used to penetrate an obscurant, the phenomenology of these obscurants (which often changes rapidly with frequency) adds an additional dimension to the problem [7]. Figure 1 - THz Imagery as related to the Frequency Spectrum. Active THz imaging systems are primarily based on highly coherent continuous-wave illumination; while optical and thermal imaging systems are based on multimodal collection of incoherent radiation. THz imaging results in coherent effects, that significantly degrade the image quality in comparison to optical and thermal images [3] as shown in Fig. 2. Fig. 2 shows the collected THz imagery for the NEACON Challenge Problem reported in Petkie et al., 2005 [8] and Jacobs et al., 2006 [9]. Also, current advances in THz imaging have been applied to medical tissue analysis [10, 11]. Figure 2 - Visible (left) and 640 GHz image (right) of target items. Top row from left to right: cigaretts, knife, radio. Middle row, left to right, lighter, metal pipe, sunglasses, screwdriver. Bottom row: left to right, gun, putty, and wallet [9]. From: E. L. Jacobs, et al., “Concealed weapon identification using terahertz imaging sensors,” Proc. of SPIE 6212, 2006. For this paper, we are interested in the THz imagery and methods of performance assessment of the fused image E. Blasch, Z. Liu, D. Petkie, R. Ewing, G. Pomrenke, and K. Reinhardt, “Image Fusion of the Terahertz-Visual NAECON Grand Challenge Data,” IEEE National Aerospace and Electronics Conf. (NAECON), 2012.