Você está na página 1de 12
 
A 3-D Stability Analysis of Lee Harvey Oswald ... JDFSL V10N3
A 3-D STABILITY ANALYSIS OF LEE HARVEYOSWALD IN THE BACKYARD PHOTO
Srivamshi Pittala, Emily Whiting, Hany Farid
Department of Computer ScienceDartmouth College
{
Srivamshi.Pittala.GR, Emily.JW.Whiting, Hany.Farid
}
@dartmouth.edu
ABSTRACT
Fifty years have passed since the assassination of U.S. President Kennedy. Despite the long passageof time, it is still argued that the famous backyard photo of Oswald, holding the same type of rifleused to assassinate the President, is a fake. These claims include, among others, that Oswald’s posein the photo is physically implausible. We describe a detailed 3-D stability analysis to determineif this claim is warranted.
Keywords
: Digital Forensics
1. INTRODUCTION
Shortly after President John F. Kennedy’s as-sassination on November 22nd, 1963, Lee Har-vey Oswald was arrested and charged with thecrime. Because Oswald was killed before histrial, we never heard from Oswald a full ac-counting of the details of the assassination.With so many unanswered questions of thatfateful day it is not surprising that numeroustheories have been proposed to fill in the miss-ing gaps. It has long been argued, for example,that Oswald did not act alone, but rather waspart of a larger criminal conspiracy involving avariety of government, international, or crimi-nal groups. These theories point to purportedinconsistencies in the events of November 22ndand in the evidence collected against Oswald.One such example is a photograph
1
of Oswaldin his backyard (States, 1964) holstering a pis-tol and holding a rifle in one hand and Marxistnewspapers in the other, Figure 1(a).This photo was particularly damning becauseit showed Oswald holding the same type of rifle
1
Although there are three backyard photos, taken inseeming succession, we focus our analysis on only one of the backyard photos.
that was used to assassinate President Kennedy.At the time of his arrest, Oswald claimed thatthis photo was fake. In addition, it has longbeen argued that the lighting and shadows inthe photo are physically implausible, that Os-wald’s facial features are inconsistent with otherphotos of him, that the size of the rifle is incon-sistent with the known length of that type of rifle, and that Oswald’s pose is physically im-plausible (it appears as if he is standing off bal-ance). The Warren Commission (States, 1964)and the House Select Committee on Assassina-tions (States, 1979) investigated claims of phototampering and concluded that they were un-warranted. More recent studies have also re-futed the claim that the lighting and shadows inthe photo are inconsistent (Farid, 2009a, 2010).This earlier work, however, did not address theclaims that Oswald’s pose is physically implau-sible. We describe the construction of a phys-ically plausible 3-D model of Oswald that in-cludes an anatomically-based mass distribution.This model allows to determine if Oswald’s poseis physically plausible.c
 2015 ADFSL Page 87
 
JDFSL V10N3 A 3-D Stability Analysis of Lee Harvey Oswald ...
2. RELATED WORK
The past decade has seen the development of many different and complementary approachesto analyzing a photo for evidence of manipu-lation (Farid, 2009b; Rocha, Scheirer, Boult,& Goldenstein, 2011). These techniques oper-ate on the assumption that manipulation willdisrupt some statistical, geometric, or physicalproperty in an image. Manipulation is exposedby quantifying and detecting these perturba-tions. For example, format-specific analyses ex-ploit artifacts that are introduced when a JPEGimage is compressed multiple times (A. Popescu& Farid, 2005; Mahdian & Saic, 2009; Kee,Johnson, & Farid, 2011). Pixel-based analysesdetect low-level correlations that are introducedby cloning (Fridrich, Soukal, & Lukas, 2003;Pan & Lyu, 2010), re-sizing (A. C. Popescu& Farid, 2005a), or non-linear filtering (Lin,Wang, Tang, & Shum, 2005). Sensor-basedanalyses can detect inconsistencies in chromaticaberrations (Johnson & Farid, 2006), color fil-ter array interpolation (A. C. Popescu & Farid,2005b; Kirchner, 2010), or sensor noise (Chen,Fridrich, Goljan, & Lukas, 2008; Fridrich, 2009).And, physically-based analyses can detect in-consistencies in reflections (O’Brien & Farid,2012), lighting (Johnson & Farid, 2005, 2007;Kee & Farid, 2010a; de Carvalho, Riess, An-gelopoulou, Pedrini, & de Rezende Rocha,2013), or shadows (Zhang, Cao, Zhang, Zhu,& Wang, 2009; Kee, O’Brien, & Farid, 2013).Physically-based methods are attractive be-cause they are applicable in low quality andlow resolution images. Within these physically-based techniques, those that can analyze 3-D scene properties are particularly attractivesince they offer a richer source of informationthan 2-D image-based techniques. 3-D stabil-ity analysis, for example, is a fundamental op-eration in engineering (Hibbeler, 2010). Be-yond design of mechanical systems, the notionof static balance has seen broad use in com-puter graphics and machine vision for anal-ysis of 3-D geometry and scene understand-ing. Static equilibrium constraints have beenincorporated into optimization techniques forobjectives such as believability of characterposes (Shi et al., 2007) and balance of 3-D-printed prototypes (Pr´evost, Whiting, Lefeb-vre, & Sorkine-Hornung, 2013). Interactive de-sign applications apply static balance to ar-chitectural structures (Whiting, Ochsendorf,& Durand, 2009; Vouga, obinger, Wallner,& Pottmann, 2012), structurally-sound furni-ture (Umetani, Igarashi, & Mitra, 2012), orsketch-based input (Derouet-Jourdan, Bertails-Descoubes, & Thollot, 2010). In the realm of 3-D scene analysis, static equilibrium has beenused as a prior in digital reconstruction, for ex-ample, to generate physically plausible urbanscenes from images (Gupta, Efros, & Hebert,2010), reconstruct missing information in pointcloud data (Shao et al., 2014), or infer orienta-tion of man-made objects (Fu, Cohen-Or, Dror,& Sheffer, 2008).Leveraging this earlier work in 3-D stabilityanalysis, we describe a new physically-based 3-D forensic analyses for analyzing the 3-D sta-bility of a person or object from a single imageand apply this analysis to the famous Oswaldbackyard photo. Digital 3-D models such asthat used here have been widely used in var-ious applications from aerospace, automotive,defense and military systems, medical systems,and product design (Duffy, 2008). Unlike life-like mannequins, digital models offer more flex-ibility in adjusting the location, pose, height,proportions, and weight of the model. In ad-dition, a digital reconstruction allows for moreflexibility in adjusting the location and type of light source, as well as the type of camera.
3. 3-D MODEL
We begin by fitting an anatomically plausi-ble 3-D model to Oswald in the backyardphoto shown in Figure 1(a). We start witha 3-D articulated human body of a genericmale (
www.makehuman.org
) of the same height(1
.
75m) as Oswald (States, 1964). The 3-Dmodel of the body used for our analysis has twocomponents: skeleton and skin. The skeleton isan armature representing major bones found inthe human body (spine, hands, arms, legs, feet,Page 88 c
 2015 ADFSL
 
A 3-D Stability Analysis of Lee Harvey Oswald ... JDFSL V10N3(a)(b)Figure 1: Shown are (a) the Oswald backyardphoto and (b) a rendering of our 3-D model.etc.) The bones are connected to each other by joints and have a rotational degree of freedomabout the joint. The skin is a textured meshof triangles that overlays the skeleton. A sim-ple colored Lambertian material is used for theskin. Although a generic 3-D head would havesufficed for our stability analysis, we employeda custom-built 3-D model generated from pho-tographs of Oswald (Farid, 2009a).The 3-D model can be realistically posedby simply rotating each body part about its joint. This manipulation respects the underly-ing physiological constraints of the human body(e.g., knees are hinge joints and an arm or legcannot be made longer or shorter). This ma-nipulation was performed in Blender, a freelyavailable 3-D rendering engine. When movinga particular body part, the skin was deformedwith the armature using the built in skinningtool in Blender, which applies parenting con-straints to bones. The articulated model withskeleton (yellow) and skin (black) is shown in aneutral pose in Figure 3(a).This articulated model was placed on aground plane with the feet making direct con-tact with the ground. A virtual camera wascreated that matched the camera used to pho-tograph Oswald – an Imperial 620 Duo Lenscamera with 35mm focal length and 36mm sen-sor size (States, 1964). We manually adjustedthe position and pose of the model’s head, torso,arms, and legs to match Oswald’s appearancein Figure 1. In the final configuration shownin Figure 1(b), the camera is 3
.
4m from the 3-D model and 1m above the ground plane. Thematched pose is shown in Figure 3(b). Becauseof the inherent ambiguity in constructing a 3-Dmodel from a single 2-D image, there are a num-ber of 3-D geometries that are consistent withthe 2-D image. The 3-D model constructed hereand the subsequent analysis show that there ex-ists a consistent and plausible 3-D scene geom-etry.Because we are interested in analyzing the 3-D stability of Oswald’s pose, it is necessary toadd anatomically plausible masses to the skele-ton. We used an anatomically-based model of human body mass distribution (DeLeva, 1996).c
 2015 ADFSL Page 89

Recompense a sua curiosidade

Tudo o que você quer ler.
A qualquer hora. Em qualquer lugar. Em qualquer dispositivo.
Sem compromisso. Cancele quando quiser.
576648e32a3d8b82ca71961b7a986505