`____________________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`____________________
`
`SAMSUNG ELECTRONICS CO., LTD. AND
`SAMSUNG ELECTRONICS AMERICA, INC.
`Petitioner
`
`v.
`
`IMAGE PROCESSING TECHNOLOGIES LLC,
`Patent Owner
`____________________
`
`CASE IPR2017-00357
`Patent No. 8,989,445
`____________________
`
`
`
`PATENT OWNER IMAGE PROCESSING TECHNOLOGIES LLC’S
`PRELIMINARY RESPONSE PURSUANT TO 37 C.F.R. § 42.107
`
`
`
`
`
`TABLE OF CONTENTS
`
`I.
`
`Introduction .................................................................................................... 1
`
`II. Overview of the ’445 Patent .......................................................................... 1
`
`A.
`
`B.
`
`Person of Ordinary Skill in the Art ....................................................... 6
`
`Claim Construction ............................................................................... 7
`
`III. Legal Standards ............................................................................................. 7
`
`IV. No Review Should be Instituted for Claims 1, 4, 6, 9, 18, 24, 25 and 27 11
`
`A.
`
`Petitioner Has Not Shown that a POSA Would Have Selected and
`Combined the Asserted References to Arrive at the Claimed
`Invention .............................................................................................. 12
`
`1. Ground 1: Petitioner Has Not Shown that a POSA Would Have
`Combined Gilbert and Brady ........................................................ 12
`
`2. Ground 2: Petitioner Has Not Shown that a POSA Would Have
`Combined Gilbert and Hashima ................................................... 15
`
`3. Ground 3: Petitioner Has Not Shown that a POSA Would Have
`Combined Hashima and Brady ..................................................... 18
`
`4. Grounds 1 and 2: Petitioner Has Not Shown that a POSA Would
`Have Combined Altan with Any of Gilbert, Brady and Hashima 22
`
`V. Conclusion .................................................................................................... 26
`
`
`
`
`
`i
`
`
`
`TABLE OF AUTHORITIES
`
`
`Cases
`Apple Inc. v. Contentguard Holdings, Inc.,
` IPR2015-00442, Paper 9 (P.T.A.B. July 13, 2015) ................................... 8, 9, 10
`
`Google, Inc. v. Everymd.com LLC,
`IPR2014-00347, Paper 9 (P.T.A.B. May 22, 2014) ............................................... 9
`
`Graham v. John Deere Co.,
`383 U.S. 1 (1966), ..............................................................................................7, 8
`
`Grain Processing v. American-Maize Prods,
` 840 F.2d 902 (Fed. Cir. 1988) ............................................................................. 11
`
`In re Magnum Oil Tools Int’l.,
` 829 F.3d 1364 (Fed. Cir. 2016) .........................................................................8, 9
`
`In re NTP, Inc.,
` 654 F.3d 1279 (Fed. Cir. 2011) ........................................................................... 11
`
`In re Omeprazole Patent Litigation,
` 536 F.3d 1361 (Fed. Cir. 2008) ........................................................................... 10
`
`InTouch Tech., Inc. v. VGo Communs., Inc.,
` 751 F.3d 1327 (Fed. Cir. 2014) ........................................................................... 11
`
`Kinetic Concepts, Inc. v. Smith & Nephew, Inc.,
` 688 F.3d 1342 (Fed. Cir. 2012) ............................................................................. 8
`
`KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398 (2007) ......................................... 10, 11
`
`Liberty Mut. Ins. Co. v. Progressive Cas. Ins. Co.,
` CBM-2012-00003, Paper 7 (P.T.A.B. Nov. 26, 2012) .....................................8, 9
`
`Ortho-McNeil Pharm. v. Mylan Labs,
` 520 F.3d 1358 (Fed. Cir. 2008) ........................................................................... 11
`
`Phillips v. AWH Corp.,
`415 F.3d 1303 (Fed. Cir. 2005) .............................................................................. 7
`
`ii
`
`
`
`Proctor & Gamble Co. v. Teva Pharm. USA, Inc.,
` 566 F.3d 989 (Fed. Cir. 2009) ........................................................................ 9, 10
`
`SAS Institute, Inc. v. ComplementSoft,
` LLC, 825 F.3d 1341 (2016) ................................................................................... 7
`
`Synopsys, Inc. v. Mentor Graphics Corp.,
` 814 F.3d 1309 (2016) ............................................................................................ 7
`
`Unigene Labs., Inc. v. Apotex, Inc.,
` 655 F.3d 1352 (Fed. Cir. 2011) ........................................................................... 10
`
`W.L. Gore & Assoc., Inc. v. Garlock, Inc.,
` 721 F.2d 1540 (Fed. Cir. 1983) ........................................................................... 12
`
`Whole Space Indus Ltd.,
` IPR2015-00488, Paper 14 (P.T.A.B. July 24, 2015) ............................................ 8
`
`Statutes
`
`35 U.S.C. § 103 ......................................................................................................7, 8
`
`
`
`
`
`
`
`
`
`iii
`
`
`
`Patent Owner Image Processing Technologies LLC (“Patent Owner”) hereby
`
`respectfully submits this Preliminary Patent Owner’s Response to the Petition filed
`
`by Samsung Electronics Co. Ltd. and Samsung Electronics America, Inc.
`
`(collectively, “Petitioner”) on November 30, 2016 in case IPR2017-00357 for
`
`review of claims 1, 4, 6, 9, 18, 24, 25 and 27 of U.S. Patent No. 8,989,445 (the
`
`“’445 patent”).
`
`I.
`
`INTRODUCTION
`
`The Board should not institute review because the Petition fails to establish a
`
`reasonable likelihood that the Petitioner would prevail with respect to any of the
`
`challenged claims.
`
`As to each of Grounds 1, 2 and 3, Petitioner fails to show that a POSA
`
`would have combined the asserted references Gilbert, Brady and Hashima (claims
`
`1, 4, 6, 9, 24, 25 and 27) and additionally Altan (claim 18) to arrive at the claimed
`
`subject matter.
`
`The Board should decline to institute an inter partes review of claims 1, 4,
`
`6, 9, 18, 24, 25 and 27 of the ’445 patent.
`
`II. OVERVIEW OF THE ’445 PATENT
`The ’445 patent is directed to efficient, real time identification and
`
`localization of a wide range of moving objects using histograms. E.g., ’445 patent,
`
`Ex. 1001 at 1:37–42; 9:17–24; 16:48–56. The inventor developed a system that
`
`1
`
`
`
`can track a target object using one or more characteristics, such as velocity, color,
`
`direction, etc. E.g., Ex. 1001 at 25:25–37; 25:61–26:3. Tracking techniques
`
`known at the time of the invention of the ’445 patent were inadequate because, for
`
`example, they were memory intensive, limited in terms of the information obtained
`
`about an object, could not provide information in real time, used complex
`
`algorithms for computing object information, or were designed to detect only one
`
`type of object. E.g., Ex. 1001, 1:43–2:9; 2:44–3:17.
`
`The ’445 patent overcame the limitations of the prior art through a number
`
`of novel techniques, including generating histograms of multiple pixel parameters
`
`being detected, with the aid of classifiers that enable only data having selected
`
`classification criteria to possibly be included in the histograms; providing a
`
`validation unit that processes multiple items of classification information from
`
`different histogram formation units in parallel to determine whether a
`
`corresponding histogram formation block will utilize data for a particular pixel in
`
`forming its own histogram; and using histograms to adjust a size of a tracking box.
`
`E.g., Ex. 1001 at 18:14–17; 18:49–55; 21:50–22:5; 24:4–25:4.
`
`Figure 11 of the ’445 patent shows a histogram processor with multiple
`
`histogram formation blocks 24–29 that are interconnected by a bus 23.
`
`2
`
`
`
`
`
`Figure 13, referring to histogram formation block 25 and validation unit 31
`
`of Figure 11, shows a histogram formation unit with a classifier 25b. The classifier
`
`has registers that permit classification criteria to be individually selected: “[b]y
`
`way of example, register 106 will include, in the case of speed, eight registers
`
`numbered 0–7. By setting a register to ‘1’, e.g., register number 2, only data that
`
`meet the criteria of the selected class, e.g., speed 2, will result in a classification
`
`output of ‘1’.” Ex. 1001 at 18:23–27.
`
`3
`
`
`
`
`
`The number of registers can vary depending on the domain of the classifier.
`
`E.g., Ex. 1001 at 18:33–45. For example, “[t]he classifier associated with
`
`histogram formation unit 24 preferably has 256 registers, one register for each
`
`possible luminance value of the image.” Ex. 1001 at 18:33–35. The interaction
`
`between the classifiers and the validation units in connection with histogram
`
`formation is significant. In particular, “[t]he output of each classifier is
`
`communicated to each of the validation blocks 30-35 via bus 23, in the case of
`
`histogram formation blocks 28 an[d] 29, through combination unit 36,” and
`
`“[v]alidation units 30–35 receive the classification information in parallel from all
`
`classification units in histogram formation blocks 24–29.” Ex. 1001 at 18:45–48.
`
`4
`
`
`
`Further, each validation unit generates a validation signal that determines “for each
`
`incoming pixel, whether the histogram formation block will utilize that pixel in
`
`forming it [sic] histogram.” Ex. 1001 at 18:51–55.
`
`The image processing system of the ’445 patent can be used to track a target,
`
`for example, a person’s face. E.g., Ex. 1001 at 22:6–33. The system may form x-
`
`axis and y-axis projection histograms using, for example, pixels whose brightness
`
`has undergone significant variation (DP=1) relative to a prior image frame. E.g.,
`
`Ex. 1001 at 10:48–63; 22:35–57. In this example, the center (V) of the face is
`
`located within an area whose boundaries are computed based on the locations of
`
`peaks in the histograms, and may correspond to the pixel position at the center of
`
`the bounded area. E.g., Ex. 1001 at 22:63–23:25.
`
`5
`
`
`
`
`
`Person of Ordinary Skill in the Art
`
`A.
`For purposes of this inter partes review, Patent Owner submits that a person
`
`of ordinary skill in the art (or “POSA”) in 1996 (the foreign priority date of the
`
`’445 patent) would be someone with an undergraduate degree in electrical
`
`engineering or image processing or a related field, followed by at least two years of
`
`graduate coursework and also at least early-stage thesis research, in digital image
`
`processing. The requisite knowledge and experience would have been acquired,
`
`for example, by someone who had completed all coursework in a two year
`
`master’s program focused on digital image processing, along with at least some
`
`thesis research qualifying towards a degree in such a program.
`
`6
`
`
`
`B. Claim Construction
`Petitioner proposes that the claims be construed pursuant to the standard in
`
`Phillips v. AWH Corp., under which “the ordinary and customary meaning of a
`
`claim term is the meaning that the term would have to a person of ordinary skill in
`
`the art in question at the time of the invention, i.e., as of the effective filing date of
`
`the patent application.” Phillips v. AWH Corp., 415 F.3d 1303, 1313 (Fed. Cir.
`
`2005).
`
`Patent Owner agrees that the Phillips standard should apply for purposes of
`
`this inter partes review.
`
`III. LEGAL STANDARDS
`Petitioner has the burden to show that it is likely to prevail as to at least one
`
`claim of the ’445 patent. 35 U.S.C. § 314. The Board may decline to institute the
`
`petition as to any claim for which the Board determines that Petitioner has not
`
`shown it is likely to prevail. SAS Institute, Inc. v. ComplementSoft, LLC, 825 F.3d
`
`1341, 1352 (Fed. Cir. 2016); Synopsys, Inc. v. Mentor Graphics Corp., 814 F.3d
`
`1309, 1316–17 (Fed. Cir. 2016).
`
`All three of Petitioner’s Grounds rely on obviousness combinations. To
`
`make a prima facie showing of obviousness for a challenged claim under 35 U.S.C.
`
`§ 103, the Petition must, among other requirements, fulfill the requirements set
`
`forth in Graham v. John Deere Co., 383 U.S. 1 (1966), including demonstrating
`
`7
`
`
`
`that the cited references, in combination, disclose each element of the claim. In re
`
`Magnum Oil Tools Int’l., 829 F.3d 1364, 1376 (Fed. Cir. 2016); see Apple Inc. v.
`
`Contentguard Holdings, Inc., IPR2015-00442, Paper 9 at 12 (P.T.A.B. July 13,
`
`2015).
`
`A legal conclusion of obviousness must be based on a factual background
`
`developed by consideration of each of the following factors: (1) the scope and
`
`content of the prior art; (2) the differences between the prior art and the claims at
`
`issue; and (3) the level of ordinary skill in the art at the time of the invention. See
`
`35 U.S.C. § 103(a); Graham, 383 U.S. 1 at 17–18. Without exception,
`
`consideration of every factor in the Graham framework is mandatory. Kinetic
`
`Concepts, 688 F.3d at 1360 (Fed. Cir. 2012); see Liberty Mut. Ins. Co. v.
`
`Progressive Cas. Ins. Co., CBM-2012-00003, Paper 7 at 3 (Order) (P.T.A.B. Nov.
`
`26, 2012).
`
`In particular, an obviousness analysis must identify the difference(s)
`
`between the claim and the prior art. Kinetic Concepts, Inc. v. Smith & Nephew,
`
`Inc., 688 F.3d 1342, 1360 (Fed. Cir. 2012) (“Indeed, courts must consider all of the
`
`Graham factors prior to reaching a conclusion with respect to obviousness.”);
`
`Whole Space Indus Ltd., IPR2015-00488, Paper 14 at 15 (P.T.A.B. July 24, 2015)
`
`(differences between the prior art and the claims at issue is one of the fundamental
`
`factual inquiries underlying an obviousness analysis); Google, Inc. v. Everymd.com
`
`8
`
`
`
`LLC, IPR2014-00347, Paper 9 at 25 (P.T.A.B. May 22, 2014) (“Rather,
`
`Petitioners’ summaries, quotations, and citations from both references, with
`
`Belanger’s figures, place the burden on us to . . . identify any differences between
`
`the claimed subject matter and the teachings of Shah and Belanger.”) (emphasis
`
`added); Liberty Mut., CBM-2012-00003, Paper 7 at 2–3 (“Differences between the
`
`claimed invention and the prior art are a critically important underlying factual
`
`inquiry for any obviousness analysis.”).
`
`Petitioner also has the burden to show whether there would have been a
`
`motivation or reason to combine the asserted prior art, and whether the proposed
`
`combination would render the patented claims obvious. In re Magnum Oil Tools,
`
`829 F.3d at 1376.
`
`A petition must provide an explicit rationale to make proposed modifications
`
`to or combinations of the prior art references, despite the differences between the
`
`claimed invention and the prior art, without relying on the patent disclosure itself.
`
`Apple Inc. v. Contentguard, Paper 9 at 15; see also Proctor & Gamble Co. v. Teva
`
`Pharm. USA, Inc., 566 F.3d 989, 995 (Fed. Cir. 2009).
`
`A petition must also explain why a person of ordinary skill in the art would
`
`simultaneously make multiple changes and implementation choices to arrive at a
`
`particular invention. Apple Inc. v. Contentguard, Paper 9 at 16–17 (“[W]e are not
`
`persuaded that the Petition sufficiently explains why a person of ordinary skill
`
`9
`
`
`
`would simultaneously make all of the many particular proposed changes and
`
`implementation choices”) (internal citations omitted). Even if individual
`
`modifications or choices were obvious, a petition must explain why making all of
`
`the changes at once would be obvious. Id. at 16–17 (“[T]he mere fact that
`
`individual changes might have been obvious does not make doing all of the
`
`changes at once obvious.”).
`
`Most inventions rely on known building blocks, so Petitioner must show that
`
`a POSA would both select and combine the building blocks “in the normal course
`
`of research and development to yield the claimed invention.” Unigene Labs., Inc.
`
`v. Apotex, Inc., 655 F.3d 1352, 1360 (Fed. Cir. 2011) (citing KSR Int’l Co. v.
`
`Teleflex Inc., 550 U.S. 398, 421 (2007)) (emphasis added). It is important to
`
`identify a reason and motivation that would have prompted a POSA to combine the
`
`prior art elements in the way claimed in the challenged patent, to achieve the
`
`invention. Proctor & Gamble Co., 566 F.3d at 994; KSR, 550 U.S. at 418–19.
`
`“Obviousness requires more than a mere showing that the prior art includes
`
`separate references covering each separate limitation in a claim under
`
`examination.” Unigene Labs, 655 F.3d at 1360 (citing KSR, 550 U.S. at 418). The
`
`lack of a technological obstacle to combining references, in and of itself, does not
`
`justify a finding of obviousness. See In re Omeprazole Patent Litigation, 536 F.3d
`
`1361, 1380-81 (Fed. Cir. 2008). A reason for combining disparate prior art
`
`10
`
`
`
`references is critical and should be made explicit. InTouch Tech., Inc. v. VGo
`
`Communs., Inc., 751 F.3d 1327, 1351 (Fed. Cir. 2014) (citing KSR, 550 U.S. at
`
`418).
`
`Hindsight analysis is inappropriate; obviousness must be measured “at the
`
`time the invention was made.” Ortho-McNeil Pharm. v. Mylan Labs, 520 F.3d
`
`1358, 1364 (Fed. Cir. 2008) (emphasis in original). The Petitioner must not use
`
`the patent as a roadmap. In re NTP, Inc., 654 F.3d 1279, 1299 (Fed. Cir. 2011)
`
`(citing Grain Processing v. American-Maize Prods, 840 F.2d 902, 907 (Fed. Cir.
`
`1988)); see also KSR, 550 U.S. at 421.
`
`IV. No Review Should be Instituted for Claims 1, 4, 6, 9, 18, 24, 25 and 27
`The Petition is deficient because it impermissibly shifts to the Board the
`
`burden of figuring out how and why the references would be combined in the
`
`manner asserted by the Petition, and in what manner the POSA would achieve the
`
`claimed invention through the asserted combination. As shown below, Petitioner
`
`does not explain how a POSA would have combined the references to arrive at the
`
`claimed invention or why a POSA would have been motivated to select and
`
`combine the references.
`
`11
`
`
`
`A.
`
`Petitioner Has Not Shown that a POSA Would Have Selected and
`Combined the Asserted References to Arrive at the Claimed
`Invention
`
`Petitioner has not shown that a POSA would have selected and combined the
`
`asserted references as set forth in the Petition to arrive at the claimed invention.
`
`The Petition fails to analyze the references as a whole, and therefore ignores
`
`disclosures in Gilbert, Brady, Hashima and Altan that would lead a POSA away
`
`from the invention of the ’445 patent. This failure to address the references as a
`
`whole by avoiding any discussion of these problematic disclosures is another
`
`reason why the petition should be denied. W.L. Gore & Assoc., Inc. v. Garlock,
`
`Inc., 721 F.2d 1540, 1548, 1550 (Fed. Cir. 1983) (a prior art reference must be
`
`considered in its entirety, i.e., as a whole, including portions that would lead away
`
`from the claimed invention).
`
`1. Ground 1: Petitioner Has Not Shown that a POSA Would
`Have Combined Gilbert and Brady
`
`Petitioner has not shown that a POSA would have looked to apply the
`
`disclosure of Brady to the disclosure of Gilbert in order to arrive at the claimed
`
`subject matter of the ’445 patent, as proposed in Ground 1. The references are
`
`directed to different objectives. Gilbert is directed to tracking a single target, or
`
`possibly two targets in separate tracking windows. Ex. 1005 at 48. Gilbert’s
`
`system includes a movable optical mount used to follow a target. Ex. 1005 at 47.
`
`In contrast, Brady is directed to monitoring vehicle traffic by switching between a
`
`12
`
`
`
`plurality of stationary cameras situated over a plurality of roadways. Ex 1007 at
`
`3:61-66. Brady is also concerned with automatic tracking of all vehicles in an
`
`image, and criticizes the prior art for being unable to do so. Ex. 1007 at 2:23–27;
`
`11:14–19.
`
`Additionally, the systems disclosed by the references operate in different
`
`ways that are incompatible. Gilbert’s objective is to identify and track objects in
`
`complex and changing backgrounds. Ex. 1005, Abstract. Gilbert discloses the
`
`determination of whether a given pixel belongs to a target, a background or a
`
`plume region based on intensity value. Ex. 1005 at 50. However, Gilbert does not
`
`seek to identify different types (or classes) of targets. Instead, Gilbert depends on
`
`the assumption that the target has some video intensities that are not contained in
`
`the immediate background. Ex. 1005 at 2. In contrast, Brady’s tracking is
`
`performed against a static background (roadways), and the tracking algorithms are
`
`specifically tailored to roadway backgrounds, by, for example, assuming that
`
`regions of interest are truck-sized trapezoids, an assumption that depends on the
`
`position of the static camera above the roadway, looking down at an angle at the
`
`rear of passing vehicles (see Ex. 1007, Figs. 1 & 5 and 7:35–45). For instance, in
`
`order to identify the class of vehicle, each “vehicle class is assigned a set of
`
`appropriately sized and shaped regions.” Ex. 1007 at 7:39–42. Candidate regions
`
`of interest are sized according to vehicle class, e.g., car sized or truck sized
`
`13
`
`
`
`trapezoids. Ex. 1007 at 7:46–54. Defining candidate regions of appropriate size
`
`would not be possible in Gilbert because there are no static points of reference
`
`such as lane boundaries and no known or predetermined vehicle shape. Without an
`
`appropriate candidate region, edgel values and fuzzy set theory would have limited
`
`utility, if any, in Gilbert’s system. Similarly, candidate regions can be dynamically
`
`generated by “prior calibration of the scene” in Brady, Ex. 1007 at 7:42–45, which
`
`would not be possible in the complex and non-stationary background of Gilbert.
`
`Petitioner has also not shown a credible reason why a POSA would combine
`
`Gilbert with Brady. The fact that both references make some use of histograms,
`
`and find a target center in some manner, Petition at 38, does not mean that a POSA
`
`would seek to combine the disclosures of the two references. Petitioner asserts that
`
`Brady’s alleged use of edgel intensity would benefit Gilbert, but fails to address
`
`that Brady uses either pre-defined target shapes such as truck-shaped trapezoids or
`
`requires training using a known background (Ex. 1007, Figs. 1 & 5 and 7:35–45),
`
`which would not be possible when tracking an object in the sky as Gilbert does.
`
`Petitioner also asserts that the alleged disclosure of intermittent processing
`
`(skipping frames) of Brady would benefit Gilbert. However, tracking a fast-
`
`moving rocket in a changing background would facially not benefit from skipping
`
`the processing of frames. While Petitioner seeks to avoid this problem by asserting
`
`that skipping frames would be helpful when tracking a rocket that is not moving (a
`
`14
`
`
`
`rocket before launch), Petition at 40, it is precisely in this easy-to-track case that
`
`Petitioner has not shown a motivation for a POSA to seek an alleged improvement
`
`in tracking capability.
`
`Therefore, because Gilbert and Brady are directed towards different
`
`objectives and operate in incompatible ways, and because Petitioner has not
`
`offered a credible reason to combine the references, Petitioner has not shown that a
`
`POSA would have combined these references as asserted in the Petition.
`
`2. Ground 2: Petitioner Has Not Shown that a POSA Would
`Have Combined Gilbert and Hashima
`
`Petitioner has not shown that a POSA would have looked to apply the
`
`disclosure of Gilbert to the disclosure of Hashima in order to achieve the claimed
`
`subject matter of the ’445 patent, as proposed in Ground 2.
`
`Petitioner’s reasons for combining Hashima and Gilbert are generally
`
`supported by superficial analysis supported by conclusory statements, including
`
`that Hashima and Gilbert “are not only in the same field of endeavor but also are
`
`directed to very similar systems and processes” and that “[a] POSA would have
`
`reasonably expected that such a combination would be successful in that it would
`
`have resulted in a more efficient computation system.”1 See, e.g., Petition at 60–
`
`61. The fact that both references have video input, make some use of histograms
`
`1 Elsewhere, for example, the Petition repeats an conclusory assertion that “[e]ach
`compliments [sic] the other such that strong disclosures of one make up for
`weaknesses in the other.” Petition at 58, 77, 81.
`
`15
`
`
`
`and track a target does not mean that a POSA would combine the references, and
`
`as shown below a POSA would have recognized the differences between the
`
`references and therefore not been motivated to select and combine them.
`
`In fact, the references are directed to different objectives. For instance,
`
`Gilbert is directed to military applications where a target is tracked from far away
`
`and is not specifically limited to any particular type of object. Hashima relates to
`
`tracking a target mounted on an object to enable a nearby robot to grip the object,
`
`and Hashima requires a predefined target with a specific and known three-
`
`dimensional shape (e.g., Ex. 1006 at Figures 2, 4), for example, a single triangle
`
`within a circle (Figure 6), two triangles within a circle (Figures 11 and 12), or a
`
`combination of both (Figure 13).
`
`The references disclose different, incompatible manners of operation. For
`
`instance, Gilbert relies on the target intensity being different than the intensity of
`
`the immediate background, Ex. 1005 at 2, whereas Hashima relies on knowing the
`
`exact three-dimensional shape of the target in advance, including that the target
`
`includes, for example, a triangle that rests on a post and thus is offset from a
`
`circular background of a contrasting color, Ex. 1006 at Figures 2 & 4 and 7:66-
`
`8:17, and uses that knowledge to track the target in three-dimensional space in
`
`order to guide the robot towards the target.
`
`16
`
`
`
`Petitioner has not shown a credible reason why a POSA would attempt to
`
`combine the two references. Petitioner asserts that Gilbert’s alleged use of
`
`histograms in multiple domains is advantageous over Hashima because the system
`
`would have a higher likelihood of successfully recognizing a target in an image
`
`frame. Petition at 61. However, Petitioner does not point to any recognized or
`
`obvious need for a higher likelihood of successful recognition in Hashima’s
`
`system, especially given that the target of Hashima is predefined in detail and
`
`contains an object with known contrasting colors, shapes, including part of the
`
`target being offset from the rest of the target in a third dimension. In fact, the
`
`combination would increase the computational burden on Hashima’s system rather
`
`than being more efficient. This would obviously be problematic for an attempt to
`
`grip a moving object or spacecraft in real-time. See Ex. 1006 at Fig. 41, 1:19-22,
`
`6:24-25, 20:30-65. In this regard, Petitioner has not shown that an increase in
`
`successful recognition would outweigh the corresponding increase in computation.
`
`Therefore, because Gilbert and Hashima are directed towards different
`
`objectives and operate in an incompatible manner, and because Petitioner has not
`
`offered a credible reason to combine, Petitioner has not shown that a POSA would
`
`have sought to combine these references in the manner asserted in the Petition.
`
`17
`
`
`
`3. Ground 3: Petitioner Has Not Shown that a POSA Would
`Have Combined Hashima and Brady
`
`Petitioner has not shown that a POSA would have looked to apply the
`
`disclosure of Brady to the disclosure of Hashima in order to achieve the claimed
`
`subject matter of the ’445 patent, as proposed in Ground 3.
`
`Petitioner asserts that it would have been obvious to modify the disclosure of
`
`Hashima based on Brady in order to (1) make Hashima’s algorithm more general,
`
`(2) reduce computational burden by intermittent tracking, (3) enhance flexibility by
`
`using edgels and fuzzy set theory, and (4) improve Hashima’s center based
`
`tracking technique. Petition at 78–81.
`
`The references, however, disclose a different, incompatible manner of
`
`operation. As to reasons (1) and (3), Hashima’s system cannot be modified to
`
`include Brady’s use of edgel intensity and fuzzy set classification because pre-
`
`defined marks, including known shapes of contrasting colors on the target and that
`
`the shape of the target is known in all three dimensions, are necessary for
`
`Hashima’s system to work as intended. Hashima’s overall objective of enabling a
`
`robot to track and grip an object requires a significant degree of tracking precision.
`
`See, e.g., Ex. 1006 at 2:13–41 (criticizing the accuracy of conventional methods for
`
`calculating the position and attitude of a target); 25:44–46 (“In FIG. 55, an object 1
`
`to be gripped by a robot hand during operation of a robot is in any optional
`
`position and has any optional attitude in a three-dimensional space.”). Thus,
`
`18
`
`
`
`Hashima has sub-objectives of “recognizing and tracking a target mark while
`
`eliminating measuring error variations due to the positional relationship between
`
`the target mark and a camera” and “measur[ing] the distance up to and the attitude
`
`of an object simply with high accuracy.” Ex. 1006 at 2:54–56. To meet these
`
`objectives, Hashima requires not only a special mark, but also a robot with a
`
`moving mechanism supporting a video camera and movable with six degrees of
`
`freedom. Ex. 1006 at 2:64–3:13. As noted earlier, Brady’s system relies on
`
`stationary cameras and includes an assumption, such as that areas of interest will
`
`have a trapezoidal truck shape, based on the known, fixed camera angle and
`
`location with respect to a known road scene.
`
`To determine shifts in the attitude of the camera coordinate system (e.g.,
`
`shifts in the yaw, pitch or roll directions), Hashima detects corresponding shifts in
`
`the image of the mark, e.g., a shift of a triangular portion of the mark in relation to
`
`a circular potion of the mark. See, e.g., Ex. 1006 at 11:47–67; Figure 27. After
`
`determining actual shifts, a speed command is generated for operating the robot’s
`
`manipulator (e.g., a gripping hand). Ex. 1006 at 18:31–37.
`
`19
`
`
`
`
` Substituting Brady’s use of edgel intensity and fuzzy set classification
`
`would not satisfy Hashima’s performance requirements with respect to detecting
`
`positional shift. Thus, Hashima’s algorithm cannot be generalized based on
`
`Brady’s disclosure.
`
`Petitioner has not shown that Hashima would benefit from the disclosure of
`
`Brady. As to reason (2) (adding Brady to reduce computational load), Hashima’s
`
`algorithm does not impose any significant computational burden in the first place
`
`that would be alleviated through Brady’s alleged use of intermittent tracking. In
`
`Hashima, “[s]ince the projected histograms are calculated in a one-dimensional
`
`20
`
`
`
`domain, they can simply be calculated and the required calculations are small.
`
`Therefore, the shifts of the target mark 10 can be determined by high-speed
`
`processing. As a result, real-time data depending on the shifts can be fed back to
`
`the robot 30, allowing the camera 20 to track the target mark on a real-time basis.”
`
`Ex. 1006 at 15:29–35. Thus, it would not have been obvious or beneficial to apply
`
`Brady’s intermittent tracking to Hashima.
`
`Additionally, Brady’s use of edgel intensity and fuzzy set classification
`
`would add computational requirements and increase the computational load. In
`
`comparison, Hashima’s formation of an intensity histogram from a binary-
`
`converted image is relatively straightforward and efficient, as it must be given the
`
`application of gripping a moving object or spacecraft in real-time. See Ex. 1006 at
`
`Fig. 41, 1:19-22, 6:24-25, 20:30-65. Therefore, a POSA would not look to Brady
`
`for reducing computational burden in Hashima.
`
`As to reason (4) (improving center-based tracking), Petitioner’s assertion
`
`that Brady’s center-based tracking is more sophisticated is supported only by
`
`Petitioner’s assertion that Brady considers inertial history. Petitioner ignores the
`
`disclosures in which Hashima describes how positional shifts can be calculated
`
`rapidly and with great accuracy using the predefined mark. E.g., Ex. 1006 at
`
`23:62–24:5 (“using a target mark in the form of a triangle of a certain shape and X-
`
`and Y-histograms of the triangle image . . . a wide range of rotational angles can be
`
`21
`
`
`
`detected stably, and shifts of rotational angles can be detected accurately.
`
`Accordingly, the time required to correct positional shifts may be shortened.”);
`
`3:26–35 (“the target mark composed of th