US 7728833

I'll verify this patent against live sources and check for any litigation or CAFC activity.

Let me do one more targeted check for USPTO/PTAB or litigation records specific to this patent.

Search scope and confidence

I ran targeted searches for the exact number 7,728,833 and screened out hits for unrelated "883" patents (e.g., a cellular inter-node connectivity patent, a plant-formulation patent in Stoller v. Fine, and Voice Domain Technologies' asserted '883 — none of these is US 7,728,833). Verbatim-text corroboration came from the patent's own full text (Google Patents), the USPTO PDF, FreePatentsOnline, Justia Patents, and uspto.report.

Litigation / CAFC check (honest limitation): I found no CAFC 2026 docket entry, district-court case, IPR/PGR, or other PTAB proceeding naming US 7,728,833. I could not query PACER or the PTAB docket directly — only via web search — so treat "no litigation found" as a negative search result, not a certification of clean title.

Bibliographic summary

Note: the granted title above is the roof-structure title shown on the face of the patent. Some secondary databases (Patents-Review, citing the pre-grant publication) carry the application-stage title "Method and apparatus performing three-dimensional computer…" — same application, different title text.

Abstract (verbatim): "A method and apparatus for automatically generating a three-dimensional computer model from a 'point cloud' of a scene produced by a laser radar (LIDAR) system. Given a point cloud of an indoor or outdoor scene, the method extracts certain structures from the imaged scene, i.e., ceiling, floor, furniture, rooftops, ground, and the like, and models these structures with planes and/or prismatic structures to achieve a three-dimensional computer model of the scene. The method may then add photographic and/or synthetic texturing to the model to achieve a realistic model."

Independent claim 1 — plain language overview

Claim 1 is a method for building a 3D roof model from a point cloud, comprising these ordered steps (from the claim text):

1. Identify a point cloud data set corresponding to the target scene (e.g., an aerial LIDAR scan).
2. Remove ground points from that data set, producing a "refined point cloud data set."
3. Identify, by a computer, one or more planar regions within the refined set.
4. Group the planar regions into local planes.
5. Group the local planes into roof structure planes, thereby creating polygons — each polygon corresponding to a face of the roof structure planes.
6. Create a topology graph of the roof portion, with each node representing one of the polygons.
7. Identify one or more parametric shapes corresponding to the topology graph.
8. Generate the 3D model of the roof structure based on those parametric shapes.

The core inventive concept per the specification: rather than fitting independent planes and then manually aligning abutting roof faces, the roof is represented by a topology graph (nodes = planar faces, edges labeled orthogonal/symmetric based on projected normals), which is matched against a database of prismatic parametric shapes; the best-fit shape is then fitted to the LIDAR points (minimizing point-to-surface distance, optionally via RANSAC), keeping the roof describable by only a few parameters (e.g., a gable roof = length, width, height, two slopes).

Dependent claims 2–7 (plain language)

• Claim 2 — ground removal detail. Identify non-planar points and remove them (leaving a planar point cloud); find the plane containing the largest number of points (that plane is the ground); remove those points to form the refined set.
• Claim 3 — local-plane grouping detail. Compute a covariance matrix of points within a planar region and group them into local planes based on the matrix eigenvalues.
• Claim 4 — roof-plane grouping detail. Compute a normal for each local plane and group together local planes whose normals are substantially parallel to define the roof structure planes.
• Claim 5 — boundary definition. Use a ball-pivoting algorithm to define the boundaries of the local planes (yielding the polygons).
• Claim 6 — topology graph scope. The topology graph may be represented by two or more topological graphs (i.e., a complex graph decomposed into sub-graphs).
• Claim 7 — shape lookup. Identifying parametric shapes includes searching a parametric shapes database for shapes corresponding to the topology graph.

Claim-drafting note: because all of claims 2–7 depend from claim 1, claim 1's combination is the only independent scope. Its step 3 carries an explicit "by a computer" limitation; the remaining recited steps are not so qualified in the claim as printed, though the specification implements the whole method in software (modeling software 114, modules 902–910, Fig. 9).

Prior art cited on the face of the patent (for context)

U.S. patents/publications cited include Dimsdale/Cyra US 6,473,079 and US 2004/0051711 (integrated LIDAR imaging/modeling); Besl US 2004/0217956; SAIC "Virtual Environment capture" US 2004/0109012 / US 7,398,481; Hayano US 2004/0041805; Harris US 6,654,690; Van Mull… no — Evans & Sutherland US 4,625,289; Kaman US 5,233,541; Lange US 2005/0117215; Strider Labs US 2005/0128197; Wu US 2005/0140670; Razdan US 2005/0168460. Non-patent literature includes Bernardini et al., "The Ball-Pivoting Algorithm for Surface Reconstruction" (IEEE TVCG 1999), Gruen & Wang, "CC-Modeler" (ISPRS 1998), Brenner, "Modelling 3D Objects Using Weak CSG Primitives" (2004), Forlani et al. and Rottensteiner et al. (LIDAR building reconstruction, 2003), and Pauly, Gross & Kobbelt (2002).

Uncertainty / caveats

• No authoritative legal-status confirmation. "Expired – Lifetime" and the Aug. 19, 2025 adjusted-expiration date come from Google Patents' automated legal-status field; Google explicitly labels this an assumption, not a legal conclusion. Independently verifying maintenance-fee payments would require USPTO Patent Center/Global Dossier access.
• No litigation or PTAB record found for US 7,728,833 via search. Absence of hits is not proof of absence.
• Assignee chain (Sarnoff → SRI International) is supported by recorded reassignment entries on the patent page; the 2013 entry is characterized there as a merger.
• OCR artifacts in the fetched full text (e.g., "modeling software 119" vs. "114"; "Indoor Modelling Component 186" vs. "118"; "smoothened" vs. "smoothed") appear to be transcription noise and do not change the substance.
• Claim text above is quoted/rephrased from the patent document itself (authoritative), not from a secondary database.

Sources: https://patents.google.com/patent/[US7728833B2](/patent/US7728833B2)/en ; USPTO patent PDF (US007728833B2); https://www.freepatentsonline.com/7728833.html ; https://patents.justia.com/patent/7728833 ; https://uspto.report/patent/grant/7728833

I'll search for litigation involving US patent 7728833 across patent litigation sources.

Let me search more specifically, since several results reference different patents ending in "883."

No direct hit yet. Let me try patent-litigation-specific databases and the Unified Patents portal, plus check assignee SRI International's known suits.

Litigation Search Results: US Patent No. 7,728,833

Bottom line

No litigation involving US Patent No. 7,728,833 was found in the searches performed. Based on the sources queried (PTAB/USPTO document repositories, CourtListener/RECAP, law-firm and PACER-derived databases, and general web search), I could not identify any district-court action, ITC investigation, or Federal Circuit appeal in which US 7,728,833 was asserted or challenged.

Confirmatory detail on the patent itself

To make sure I was matching the right asset, the patent in question is:

Important caveat — do not conflate with other "883" patents

My searches repeatedly surfaced different patents whose numbers happen to end in "883," and these are not US 7,728,833. Per your instruction to interpret identifiers literally and not auto-correct, I flag them so they are not misattributed:

• US 5,563,883 ("Dynamic Channel Management and Signalling Method and Apparatus") — asserted in C-Cation Technologies, LLC v. Comcast Corp., Charter Communications, et al. (E.D. Tex.).
• Dominion Resources, Inc. v. Alstom Grid Inc. — an unrelated '883 patent concerning voltage/CVR (E.D. Pa.).
• Brocade Communications Systems, Inc. v. A10 Networks, Inc. — unrelated '883 patent.
• A cellular-network '883 patent challenged in an IPR (prior art: Centonza, Han).

None of these involve US 7,728,833, and none should be reported as litigation over it.

Confidence and limitations

• High confidence that the searches did not surface any suit over US 7,728,833.
• Moderate confidence in the completeness of this "no litigation" finding. Patent litigation histories are not always fully indexed in free sources, and an exhaustive negative requires dedicated litigation databases (e.g., Docket Navigator, Lex Machina, Darts-IP) that I could not fully query within the search limits. A prior-owner/SRI International licensing campaign that produced only demand letters or confidential settlements could also exist without a public docket.
• The patent expired in 2025, which further reduces the likelihood of any active or recent assertion.

If you need a definitive negative, the recommended next step is a targeted Docket Navigator or Lex Machina search on patent number 7,728,833 (not the truncated "'883"), and a check of SRI International's litigation history as assignee. I can draft that query set for you if helpful.

I'd rather flag this uncertainty than fabricate a case caption, jurisdiction, or docket number.

Proceedings overview

Zero (0) AIA trial proceedings are on file against US 7,728,833. The structured PTAB proceedings on file block — sourced from the USPTO Open Data Portal — returns no IPR, PGR, or CBM for this patent, and my independent web searches surfaced none. Breakdown: 0 active / 0 claims invalidated / 0 claims sustained / 0 settled / 0 institution-denied. Bottom line for a defendant: this patent has never been tested at the Board. All 7 claims stand unadjudicated, so unlike a hardened post-IPR patent, there is no estoppel record and no Board-construed claim scope — which cuts both ways (no free kill, but also no settled narrowing you must work around).

Important disambiguation flag: a live search hit for "Case IPR2017-00067, Patent 7,910,833 B2" (Superior Communications v. Voltstar Technologies) is not this patent. That is U.S. Patent 7,910,833 (a different patent, different owner, different technology — power adapters). Likewise, hits for "833" in Brazilian case numbers, Norwegian avalanche coordinates, and FEC transaction IDs are noise. Do not cite any of those as proceedings against US 7,728,833. Only US 7,728,833 / app. 11/206,665 counts.

Proceedings

None to report. There is no proceeding to list, so the per-proceeding template is intentionally empty rather than filled with fabricated numbers.

I will not invent an IPR number, petitioner, or FWD. If a proceeding exists that neither ODP nor web search has indexed, it is not public and I cannot responsibly describe it.

Strategic summary

Claim status: all 7 claims UNTESTED. No claim of 7728833 has been canceled, confirmed, or even challenged at the PTAB. Claim 1 (the sole independent claim — point cloud → ground removal → planar-region/local-plane/roof-plane grouping → topology graph → parametric-shape match → 3D roof model) remains in its as-issued form, as do dependent claims 2–7. Contrast this with the heavily-litigated "833" patent in the search results — that's a different patent, and its 7,910,833 IPR history tells you nothing about this one.

Estoppel: none to worry about, none to rely on. Because no IPR/PGR was ever instituted against this patent, § 315(e)(2) estoppel is inapplicable — no petitioner (or privy) is barred, and no prior-art ground is off the table for a would-be challenger. Conversely, a defendant cannot lean on any earlier petitioner's Board findings, because there are none. The entire prior-art landscape — including the art cited on the face of the patent (Dimsdale/Cyra '079, SAIC Virtual Environment Capture, Besl, Hayano, Harris '690) and the non-patent literature (Bernardin ball-pivoting 1999; Gruen & Wang CC-Modeler 1998; Brenner Weak CSG Primitives 2004; Forlani 2003; Rottensteiner 2003; Pauly/Gross/Kobbelt 2002) — is fully available as IPR grounds.

Pattern signals: absent. No repeat petitioner, no serial IPR campaign, no Unified Patents (or similar defensive aggregator) challenge appears in the record for this patent. That is a meaningful signal in itself: a patent asserted against deep-pocketed defendants (the EagleView/Pictometry/Hover/Xactware ecosystem that cites this patent extensively) would typically draw an IPR, yet 7728833 has attracted none in ~15 years of post-issuance life. The most likely explanations are (a) the patent was never asserted in a campaign that justified IPR spend, and/or (b) it is now expired (Google's automated adjusted-expiration date of 2025-08-19), so prospective challengers have little incentive to file. That expiration is the single most important strategic fact here.

Forward-looking CAFC: No Federal Circuit appeal exists because there is no PTAB decision to appeal. Nothing to link on CourtListener.

Recommended next steps

• If you're a defendant and the patent is expired (or expiring): any ongoing assertion of 7728833 for future infringement is weak-to-nonexistent for conduct after the expiration date; your exposure is limited to the pre-expiration damages window. Confirm the expiration/reinstatement status via USPTO Patent Center and Global Dossier (the Google date is an assumption, not a legal conclusion) before relying on this.
• If pre-expiration back-damages are still on the table: because there is no § 315(e) estoppel, you are free to file an IPR on the best art you can find — and there is no prior petitioner's failed effort that the Board would feel bound by under General Plastic / § 325(d). The patent's dependence on a single independent claim (claim 1) makes it a compact IPR target: invalidate claim 1 and all of 2–7 fall.
• If you cannot file an AIA trial (e.g., poor art, or expired patent moots the value): fall back on district-court invalidity — the cited NPL (ball-pivoting, CC-Modeler, weak-CSG primitives) is exactly the kind of printed publication that maps onto claim 1's topology-graph/parametric-shape steps.
• Verify rather than assume: I could not query PACER or the PTAB docket directly, only via web search. Treat "no PTAB activity found" as a negative search result, not a certification. Confirm via USPTO PTAB E2E and Patent Center before finalizing any defense strategy.

Sources checked: USPTO Open Data Portal (structured block, authoritative for the proceeding list); Google Patents US7728833B2; USPTO patent PDF US007728833B2; FreePatentsOnline 7728833; Justia Patents. No PTAB E2E or CourtListener opinion exists to link for this patent.

I'll research the assignment record for US 7,728,833 using the USPTO Assignment Center and cross-references.

Found a merger assignment record. Let me dig deeper into the specific reel/frame entries and the original inventor-to-Sarnoff assignment.

I hit the search-step ceiling before I could fully triangulate every reel/frame, but I gathered enough to reconstruct the chain with clear sourcing. Critically: the Assignment Center itself was not directly queryable through my tools — everything below comes from Google Patents' legal-events feed and Plainsite's mirror of the USPTO assignment records. Treat reel/frame where shown as sourced, and where blank as not found, not absent.

Inventors

• The application was filed Aug. 18, 2005 and assigned to Sarnoff Corporation by all four inventors, corroborated by the USPTO assignment recorded in the Nov. 9, 2005 legal event, which lists assignors "KUMAR, RAKESH, HSU, STEPHEN CHARLES, VERMA, VIVEK, SAWHNEY, HARPREET."
• Unusual pattern flagged: Hsu's residence (Sunnyvale, CA) is anomalous among three New Jersey–based co-inventors and sits in SRI International's home region (Menlo Park/Silicon Valley). That is consistent with an SRI-side contributor, but I have no evidence of individual inventor mobility, departure timing, or a post-filing exodus. The "all inventors leave within 12 months → fire-sale precursor" pattern is unclear, not present.
• Note the patent carries U.S. government rights (contract N61339-03-C-0090), which constrains any later exclusive transfer.

Original assignee

Sarnoff Corporation, Princeton, NJ (originally named on the issued patent; pre-grant assignment recorded to Sarnoff Nov. 9, 2005).

• Line of business: contract R&D laboratory — the former RCA Laboratories / David Sarnoff Research Center. Historically a defense/consumer-electronics research house (patent-heavy; the sources note it received more patents than Bell Labs in its RCA era). It did not sell consumer products embodying the '883 roof-modeling claims; its model was government/industry research contracts plus patent licensing income.
• Status: Sarnoff descended from RCA Labs, was transferred by GE to SRI International in 1987 (operated as a for-profit SRI subsidiary), renamed Sarnoff Corporation in 1997, and was "fully integrated into SRI in 2011." It is therefore no longer a separate operating entity — it was absorbed into its parent, not sold off and not bankrupt. This is an internal corporate succession, not a distress event.
• Because Sarnoff's business included patent licensing, it is technically a "non-practicing" licensor in the product sense — but it is an original R&D lab, not an NPE/secondary-market acquirer.

Assignment timeline

Two recorded events surfaced. The first is documented on Google Patents' legal-events feed; the second is documented both there and on Plainsite's USPTO mirror.

• 2005-11-09 (executed date not separately shown) / recorded 2005-11-09 — Reel not surfaced in available sources (reel/frame unknown)

  • Conveyance: Assignment (assignment of assignors' interest; "SEE DOCUMENT FOR DETAILS")
  • Assignor: Vivek Verma; Rakesh Kumar; Stephen Charles Hsu; Harpreet Sawhney (all four inventors)
  • Assignee: Sarnoff Corporation (Princeton, NJ)
  • Correspondent: not surfaced — I could not retrieve the correspondent of record for this record from the sources available; not fabricated here.
  • Context: routine inventor-to-employer assignment consummated ~3 months after the Aug. 18, 2005 filing (employer owns the application ab initio).

• 2011-02-04 (executed) / recorded 2013-09-30 — Reel 31308 / Frame 0208

  • Conveyance: Merger
  • Assignor: Sarnoff Corporation
  • Assignee: SRI International, Inc. (Menlo Park, CA; "officially listed as SRI International")
  • Correspondent: not surfaced in the source page.
  • Context: internal reorganization / parent-subsidiary merger (Sarnoff fully integrated into SRI in 2011); this is a bulk merger recording covering many Sarnoff patents, not a transfer to an asserter.
  • Date discrepancy to flag: Google Patents lists this merger reassignment as dated 2013-09-11, while Plainsite shows the recording date as 2013-09-30 for the same executed-Feb.-4-2011 merger (reel 31308/0208). These are almost certainly the same transaction reported at two different granularities (event date vs. recordation date); the prior "Patent summary" section's "Sep. 11, 2013" figure should be read against Plainsite's 2013-09-30, both postdating the Feb. 4, 2011 execution. Treat the execution date (2011-02-04) as the legally operative transfer date.

No other assignments, security interests, licenses, releases, corrections, or change-of-name records were found for US 7,728,833. I did not find a record for the original inventor assignment's reel/frame; that is a gap, so do not quote a reel/frame for it.

Timeline diagram

timeline
    title Ownership of US 7728833
    2004 : Priority filings by inventors
    2005 : Application filed Aug 18
         : Assigned to Sarnoff Corporation
    2011 : Sarnoff merged into SRI International
    2013 : Merger recorded at reel 31308 frame 208
    2025 : Adjusted expiration listed

NPE / troll-pattern signals

1. Shell-entity transfer — not present. The only post-inventor transfer is Sarnoff → SRI International, a merger (reel 31308/0208, executed 2011-02-04) into a nonprofit research institute. No "IP/Holdings/Ventures/Licensing" LLC, no registered-agent-service address, no single-member Delaware/Texas shell appears in the chain.

2. Known asserter in the chain — not present. Neither Sarnoff Corporation nor SRI International appears on the Acacia / Marathon / Intellectual Ventures / IPNav / Wi-LAN / Mosaid-Conversant / Vringo / Pendrell / Round Rock / MPHJ lists. SRI is a legitimate R&D institute and a regular plaintiff-defendant in ordinary IP disputes on other patents, but it is not a high-frequency NPE plaintiff of the kind RPX/Unified track, and I found no Unified/RPX listing tying this patent to such a plaintiff.

3. Repeat correspondent across the chain — unclear / not established. I could not retrieve the correspondent of record for either the 2005 assignment or the 2013 merger recording from the sources available. A single appearance would not be probative anyway; recurrence is the signal. No finding — this is a data gap, not a clean bill.

4. Cascading transfers — not present. There is exactly one downstream link (Sarnoff → SRI), executed 2011-02-04, i.e., a single corporate merger — not multiple chained LLC hops within 24 months.

5. Pre-litigation transfer — not present. I found no infringement suit naming US 7,728,833. With no suit in the record, a "within 6 months before filing" transfer cannot exist here.

6. Bankruptcy fire-sale — not present. Sarnoff was integrated into SRI by merger (2011), not liquidated in Chapter 7/11. No bankruptcy sale vehicle (cf. Kodak/Nortel/Polaroid) appears.

7. Privateering — not present. Sarnoff did license patents, but there is no NPE asserting on Sarnoff's or SRI's behalf against competitors in the record for this patent.

8. Defensive aggregator — not present. The chain terminates at SRI International, which is not RPX, AST, LOT Network, Unified Patents, or OIN. So there is no affirmative defensive-neutralization finding either.

Verdict

Insufficient data — which here is a clean-chain finding, not a no-records finding: the records that exist show only the original inventor→Sarnoff assignment and a single Sarnoff → SRI International merger (reel 31308/0208, executed 2011-02-04, recorded 2013-09-30). Zero NPE signals are present — no shell LLC, no known asserter, no cascading transfers, no pre-litigation transfer, no bankruptcy — so this is an ordinary corporate-succession chain that terminates at a nonprofit research institute, not an assertion vehicle. I decline to label it "Operating-company assertion" (SRI does not ship products embodying these claims and is not suing competitors here) and decline "Defensive / non-asserting" (SRI is not a defensive aggregator); the option set simply lacks a "clean chain / no NPE indicators" bucket, so I record it as Insufficient data with no adverse NPE indicators.

Two honest limitations: (a) I could not query USPTO Assignment Center directly with my tools — reel/frame 31308/0208 and the 2011-02-04 execution date come from Plainsite's USPTO mirror, and the correspondent of record for both links is unretrieved; (b) I could not retrieve the reel/frame for the original 2005 inventor assignment. Verify both at the Assignment Center search page:

https://assignmentcenter.uspto.gov/ (search "7728833" or application "11/206,665")

Sources used: Google Patents legal events (https://patents.google.com/patent/[US7728833B2](/patent/US7728833B2)/en); Plainsite USPTO assignment mirror, Patent Assignment 31308/208 — Sarnoff Corporation → SRI International, Inc., Merger (https://www.plainsite.org/patents/assignment.html?id=[6071118](/patent/6071118)); Sarnoff/SRI corporate history (https://www.encyclopedia.com/books/politics-and-business-magazines/sarnoff-corporation; https://ncyclopedia.thefreedictionary.com/SRI+International).

I'll search the USPTO and other sources for this specific patent and verify the prior-art references cited on its face.

Prior-Art Analysis for US 7,728,833 — Examiner-Cited References

A. Patent-identity check (USPTO / Google Patents)

Search confirms the target asset is exactly US 7,728,833 B2 ("Method for generating a three-dimensional model of a roof structure"), App. No. 11/206,665, filed 2005-08-18, granted 2010-06-01, priority to provisional 60/602,354 (2004-08-18) and provisional 60/693,718 (2005-06-24), inventors Verma/Kumar/Hsu/Sawhney, original assignee Sarnoff, current SRI International. Source: https://patents.google.com/patent/[US7728833B2](/patent/US7728833B2)/en. No similarly-numbered record was substituted.

B. Critical legal framing (do not skip)

Three honest caveats govern everything below:

1. These are the references the examiner cited and considered, and the patent issued over them. As a matter of the prosecution record, none of them was found to anticipate any claim. A § 102 rejection requires that a single reference disclose every element of a claim, arranged as in the claim. For the reasons set out in § D, none of the cited references appears — on its face — to disclose the full combination of claim 1 (particularly the topology-graph → parametric-shape-database keystone).
2. I have not performed a verbatim, element-by-element read of every reference's full specification and file-wrapper record; the analysis below is a relevance screen for potential § 102 exposure, not a validity opinion.
3. Most of these references are more probative as § 103 (obviousness) art than as § 102 anticipation art. I flag where a single reference comes closest to anticipation.

C. The cited US patent documents (15) — citation, dates, description, § 102 relevance

Group 1 — Closest references

C1. US 6,473,079 B1 — Kacyra, Dimsdale & Brunkhart — Cyra Technologies, Inc.
"Integrated system for quickly and accurately imaging and modeling three-dimensional objects."

• Filing 1998-10-23 (continuation of App. 08/638,961); granted 2002-10-29. Parent US 5,988,862 granted 1999-11-23.
• Description: a scanning laser device scans a 3D object and generates a point cloud; a first model is generated representing constituent geometric shapes of the object; a data file is output to a CAD system. (Verified: Justia/Unified Patents/FPO.)
• § 102 relevance: Closest single reference. Discloses the point-cloud→geometric-shape model→CAD pipeline (claim-1 preamble and spirit). Does not disclose (i) removing ground via largest-plane/non-planar filtering, (ii) the local-plane→roof-plane grouping, (iii) a topology graph of roof faces, or (iv) matching that graph to parametric shapes. Available as § 102(b) art (granted >1 yr before 2004-08-18). Potential anticipation: none of claims 1–7 in full; contributes to § 103 combinations.

C2. US 6,654,690 B2 — Harris Corporation
"Automated method for making a topographical model and related system."

• Filed 2001-04-05; granted 2003-11-25; also published as US 2002/0147567 A1 (2002-10-10).
• Description: from randomly spaced LIDAR elevation-vs-position data, resample to a grid; distinguish building data from terrain data using position windows; perform polygon extraction for building data (boundaries → vertices → polygons); optionally merge optical imagery onto buildings. (Verified: Google Patents/FPO.)
• § 102 relevance: Strong on the front end of claim 1 — LIDAR in, ground/terrain separated from buildings, polygons created for buildings. Published 2002-10-10, i.e., § 102(b) art (>1 yr before the 2004-08-18 priority). Does not disclose topology graphs or parametric-shape-database matching. Its terrain/building separation is by gridded position windows, not by the claim-2 “non-planar-point removal + largest-point-count plane = ground” technique, so it does not anticipate claim 2. Potential anticipation: partial as to claim 1’s planar-region/polygon concepts; no complete single-reference anticipation.

C3. US 2004/0051711 A1 — Jerry Dimsdale
"Integrated system for quickly and accurately imaging and modeling three-dimensional objects."

• Priority 1996-04-24; published 2004-03-18 (pub. of the Cyra family, sibling of C1).
• Description: same Cyra integrated laser-scan/point-cloud/geometric-model disclosure.
• § 102 relevance: § 102(e) art (US application with 1996 priority, published before the inventor’s date). Same scope limitation as C1 — no topology graph, no parametric shape database. No complete anticipation.

Group 2 — Moderately relevant (modeling / reconstruction)

C4. US 2004/0041805 A1 — Tomoaki Hayano
"Automatic generating device for 3-D structure shape, automatic generating method, program therefor, and recording medium recording the program."

• PCT filed 2001-07-31; published 2004-03-04.
• Description: automated generation of 3D structure shapes. Relevant to the general "automatically generate a 3D model" goal.
• § 102 relevance: § 102(e)/(a) art. Does not disclose the LIDAR-ground-removal or topology-graph/parametric-shape limitations. No complete anticipation; background.

C5. US 7,398,481 B2 — Science Applications International Corp. (SAIC)
"Virtual environment capture." (Family member of US 2004/0109012 A1, below.)

• Filed 2002-12-10; granted 2008-07-08.
• Description: capturing a virtual environment from scanned/point data. Relevant to the point-cloud→environment-model concept.
• § 102 relevance: § 102(e) art (filed 2002-12-10, before applicant’s date). No topology-graph/parametric-shape teaching. No complete anticipation.

C6. US 2004/0109012 A1 — Science Applications International Corp. (SAIC)
"Virtual Environment capture."

• Filed 2002-12-10; published 2004-06-10.
• Description: as C5 (parent publication).
• § 102 relevance: § 102(e)/(a). No complete anticipation.

C7. US 2005/0128197 A1 — Strider Labs, Inc.
"Probable reconstruction of surfaces in occluded regions by computed symmetry."

• Filed 2003-12-11; published 2005-06-16.
• Description: reconstructing surfaces in occluded regions using computed symmetry.
• § 102 relevance: Notably, the ’833 specification uses symmetric (S+/S–) edge labels and constrains gable-roof planes to be symmetric. Strider’s “computed symmetry” surface reconstruction is topically adjacent but is directed to occluded-surface inference, not roof-face topology graphs or fits to LIDAR roof planes. § 102(e) art. Weak; no complete anticipation of any claim.

C8. US 2005/0140670 A1 — Hong Wu
"Photogrammetric reconstruction of free-form objects with curvilinear structures."

• Filed 2003-11-20; published 2005-06-30.
• Description: photogrammetric reconstruction of free-form objects. § 102(e) art; not roof-topology specific. No complete anticipation.

C9. US 2006/0006309 A1 — Jerry Dimsdale
"Method and apparatus for high resolution 3D imaging."

• Filed 2004-07-06; published 2006-01-12 (i.e., after the 2004-08-18 priority).
• Description: high-resolution 3D imaging (point-cloud acquisition).
• § 102 relevance: Because it published after the priority date, it can only be prior art as a § 102(e) reference (US application filed 2004-07-06, before applicant’s invention). Directed to imaging/acquisition, not roof modeling. No anticipation.

Group 3 — Background / peripheral (cited for general context)

C10. US 4,625,289 A — Rockwood (Evans & Sutherland Computer Corp.) — "Computer graphics system of general surface rendering by exhaustive sampling." Filed 1985-01-09; granted 1986-11-25. Z-buffer/exhaustive-sampling surface rendering; estimates surface normals from neighboring samples. (Verified: patent PDF, FPO.) General graphics rendering — § 102(b) art but discloses nothing about point-cloud roof modeling, ground removal, planar grouping, topology graphs, or parametric shapes. No anticipation.

C11. US 5,233,541 A — Corwin et al. (Kaman Aerospace Corp.) — "Automatic target detection process." Filed 1990-08-10; granted 1993-08-03. Model-based statistical detection of targets from imaging-LIDAR images. (Verified: patent PDF, FPO.) Lidar-based but directed to target detection, not 3D roof reconstruction. No anticipation.

C12. US 2002/0041327 A1 — Hildreth (Evan) — "Video-based image control system." Filed 2000-07-24; published 2002-04-11. Image-control; peripheral. No anticipation.

C13. US 2004/0217956 A1 — Paul Besl — "Method and system for processing, compressing, streaming, and interactive rendering of 3D color image data." Filed 2002-02-28; published 2004-11-04. 3D point/color-data handling; peripheral. No anticipation.

C14. US 2005/0117215 A1 — Eric B. Lange — "Stereoscopic imaging." Filed 2003-09-30; published 2005-06-02. Stereo capture; peripheral. No anticipation.

C15. US 2005/0168460 A1 — Anshuman Razdan — "Three-dimensional digital library system." Filed 2002-04-04; published 2005-08-04. 3D digital-library management; peripheral. No anticipation.

D. Non-patent literature (6) — the most substantive art for the keystone limitations

The NPL references are, on substance, the most relevant to the inventive core of claim 1. Verified titles/dates (as listed on the ’833 face and consistent with search):

E. Bottom-line § 102 assessment for the cited references

Because all six dependent claims depend from claim 1, the only anticipation question that matters is whether a single reference discloses all of claim 1’s steps: point-cloud → ground removal → planar-region identification → local-plane grouping → roof-plane grouping into polygons → topology graph (nodes = polygons) → identify parametric shapes corresponding to that graph → generate the roof model from those shapes.

• No cited reference, on its face, discloses the full claim-1 combination. The closest approaches are split:
  • front-end (point cloud / ground-vs-building separation / polygon extraction): US 6,654,690 (Harris) and the Cyra US 6,473,079 / US 2004/0051711 family;
  • keystone (topology graph of roof faces): Gruen & Wang CC-Modeler (N2);
  • keystone (parametric/prismatic shape fitting to LIDAR): Brenner (N3), with Forlani (N4) and Rottensteiner (N5) supplying roof-segment extraction.
• Pers-claim sketch of exposure: claim 5 (ball-pivoting) is squarely met as a technique by Bernardini (N1); claim 6 (multiple/sub-graphs) is topically met by N2; claim 3 (covariance/eigenvalue local planes) is topically met by N6/N4/N5; claim 7 (shape-database lookup) has only loose analogs in N3. But none of these references alone supplies the claim-1 scaffolding these dependents require, so none is a clean § 102 anticipation.
• The realistic validity posture of the ’833 patent vis-à-vis this art is § 103 obviousness, not § 102 — i.e., whether the topology-graph→parametric-shape-database combination would have been obvious over, e.g., Harris/Cyra in view of CC-Modeler and Brenner.
• What I could not verify: the full text of several Group-2/Group-3 references and the file-wrapper/IDS to confirm which references the examiner actually applied versus merely listed. Treat the above as a relevance screen, not a validity opinion. For a definitive § 102/§ 103 determination, an element-by-element chart of each reference against each claim (with claim-construction of “topology graph” and “parametric shapes”) is required.

Sources: https://patents.google.com/patent/US7728833B2/en ; https://www.freepatentsonline.com/[7728833](/patent/7728833).html ; https://patents.justia.com/patent/7728833 ; patent PDFs via patentimages.storage.googleapis.com (US4625289, US5233541, US5988862); https://www.freepatentsonline.com/[6654690](/patent/6654690).html ; https://portal.unifiedpatents.com/patents/patent/US-[4625289](/patent/4625289)-A ; https://portal.unifiedpatents.com/patents/patent/US-[5988862](/patent/5988862)-A.

I'll ground this §103 analysis by verifying what the key cited references actually disclose, since the patent's own text characterizes them only briefly.

Good grounding. Let me verify the remaining key references (ball-pivoting, point-cloud simplification, and the Cyra/Dimsdale and Hayano patent disclosures).

Obviousness Analysis — US 7,728,833 B2 under 35 U.S.C. § 103

Builds on the prior "Patent summary," "Independent claim 1 — plain-language overview," and "Dependent claims 2–7" sections above. Those are treated as authoritative and are not repeated; the claim-element numbering used here matches that earlier breakdown.

1. Framework and controlling date

Governing law. The application was filed August 18, 2005 (priority to provisional 60/602,354, Aug. 18, 2004), i.e., before March 16, 2013. Pre-AIA §§ 102/103 therefore govern, and KSR Int'l Co. v. Teleflex Inc., 550 U.S. 398 (2007), supplies the obviousness rubric that applies to this patent's prosecution tail (granted June 1, 2010).

Critical date. Every reference relied on below predates Aug. 18, 2004, so each is prior art under pre-AIA § 102(a) against the earliest claimed priority date, and each is also § 102(b) art relative to the Aug. 18, 2005 non-provisional filing:

Note on the Aug. 18, 2004 vs. June 24, 2005 priority question. Even if some claim limitations were entitled only to the 60/693,718 (June 24, 2005) date — the later provisional — every reference above still predates it. The obviousness conclusion is therefore insensitive to how the priority question resolves. That is a useful robustness point; it means no ground below can be defeated merely by attacking priority.

Level of ordinary skill (PHOSITA). A person with a graduate degree (or equivalent experience) in photogrammetry, computer vision, remote sensing, or geomatics, with 2–3 years' experience in airborne/terrestrial LIDAR point-cloud processing and 3D building reconstruction, familiar with: terrain/DTM filtering, region growing and RANSAC plane fitting, covariance/eigenvector (PCA) local-surface estimation, polygonal surface reconstruction from point sets, and parametric/CSG building-model libraries. The patent's own specification assumes this level (it calls RANSAC, Delaunay triangulation, morphological operators, and ICP "well-known").

The patent's own admissions are significant. The specification concedes that (a) LIDAR "is increasingly becoming the modality of choice" for 3D scene data; (b) manual photogrammetric model construction was the known baseline; and (c) an entire first technique for roof extraction — "the roof can be defined by one or more independent planes… A well-known RANSAC (RANdom Sample Consensus) algorithm is used to best fit the plane to the data" — was already known, with the claimed invention described as an alternative that "simplifies the rooftop modeling process and mitigates the processing used to align the abutting edges." A stated advantage framed purely as simplification and reduced processing is the classic KSR predicate (KSR, 550 U.S. at 417, 421).

2. What the cited art actually teaches (verified against live sources)

Because the patent lists these references without substantive discussion, I re-verified the key ones. Verified content:

2.1 Forlani et al. 2003 — Building Reconstruction and Visualization from LIDAR Data

Verbatim from the paper's abstract (Semantic Scholar / Academia):

"Roofs are modeled as plane surfaces, connected along ridges and bordered by the eaves lines. Initial segmentation of terrain and non-terrain features is performed on grid data; after segmentation by region growing and region topology, classification is achieved based on a hierarchical set of aggregation rules. Areas labelled as potential buildings are further segmented in plane surfaces (roofs slopes) based on gradient orientation and plane fitting by RANSAC; a similar procedure applies to edge pixels to extract eaves lines. The topology of the roof slopes and walls is reconstructed, deriving also roof ridges and roof corners. A 3D model of the building is thereafter obtained…"

Sources: https://www.semanticscholar.org/paper/Building-Reconstruction-and-Visualization-from-Data-Nardinocchi-Forlani/268faca762f20f8197565fb42b9556fda4706dc7 ; https://www.academia.edu/48635719/Building_reconstruction_and_visualization_from_lidar_data

What this reaches: point-cloud-based building extraction; terrain/non-terrain separation (= ground removal); classification of non-building features; segmentation into planar roof slopes; RANSAC plane fitting; and — critically — reconstruction of roof topology (slopes, walls, ridges, corners). Forlani is the single closest reference to claim 1's preamble through the "roof structure planes / polygons" step.

2.2 Rottensteiner et al. 2003 — Detecting Buildings and Roof Segments by Combining LIDAR Data and Multispectral Images

Verbatim from the paper (CORE PDF) and abstract:

"A method for the automatic detection of buildings and their roof planes from LIDAR data and multispectral images… The detection of roof planes is based on a region growing algorithm applied to the LIDAR data, the seed regions detected by a grey-level segmentation of the multispectral images."

and, in the method body:

"Morphological filtering provides us with an approximation for the DTM… An initial building mask is created by thresholding the height differences between the last pulse DSM and the DTM… pixels having an NDVI above a certain threshold are erased in the building mask… large height differences between first and last pulse data indicate trees… 'homogeneous' pixels correspond to areas of locally parallel surface normal vectors, thus, they are situated in a locally planar neighbourhood. 'Linear' pixels correspond to the intersections of planes, and 'point-like' pixels are in a neighbourhood of great, but anisotropic variations of the surface normal vectors."

Sources: https://core.ac.uk/works/[6238648](/patent/6238648)/ ; https://core.ac.uk/download/14982344.pdf

What this reaches: ground/terrain separation via a morphological DTM; classification and elimination of trees and vegetation from the building mask; identification of locally-planar neighborhoods and rejection of non-planar (point-like) neighborhoods — i.e., the point-level planarity test and the removal of non-planar structures that claim 2 recites — and region-growing detection of roof planes.

2.3 Gruen & Wang 1998 — CC-Modeler: a topology generator for 3-D city models

Verbatim from the paper (ETH Research Collection PDF and ScienceDirect abstract):

"Given the data as point clouds… we present a new method for fitting planar structures to the measured sets of point clouds… The problem of fitting planar faces to point clouds is treated as a Consistent Labelling problem, which is solved by probabilistic relaxation. Once the faces are defined and the related points are determined, we apply a simultaneous least-squares adjustment in order to fit the faces jointly…"

and:

"The next step is to build the face model of the 3-D object, i.e. to determine how many faces the 3-D object has, which points define an exact face and the spatial relations of the faces… This adjustment is amended by observation equations that model orthogonality constraints of pairs of straight lines between boundary points… CC-Modeler has the ability to map images onto a 3-D object… The texture, taken from the original images, is attached to an exact face."

Sources: https://www.research-collection.ethz.ch/bitstream/handle/20.500.11850/[150916](/patent/150916)/eth-31183-01.pdf ; https://www.sciencedirect.com/science/article/abs/pii/S0924271698000112 (DOI 10.1016/S0924-2716(98)00012-4)

What this reaches — and why it matters most. CC-Modeler is a "topology generator." It builds a face model in which each roof face is a unit of the model and the "spatial relations of the faces" are explicitly computed, with orthogonality constraints between faces encoded in the adjustment. That is, in substance, claim 1's "topology graph… wherein each node represents one of the one or more polygons" — a graph of faces with labeled adjacency relations. The examiner cited it. This is the reference that most directly erodes the patent's asserted point of novelty.

2.4 Brenner 2004 — Modelling 3D Objects Using Weak CSG Primitives

Verbatim from the paper (IKG Hannover PDF):

"(Gülch and Muller, 2001…) describe an approach which is based on CSG primitives measured semi-automatically in aerial images. Buildings are modelled using a fixed number of parametric primitives like flat-, desk-, saddleback-, hipped-roof etc., which are combined by Boolean operations.… Then, the wireframe model is overlaid… and the operator can adapt the parameters… Regularity is enforced implicitly by the primitives."

"(Brenner, 1999) divides ground plans into 2D primitives… For each of the 2D primitives, an optimal 3D primitive is selected from a fixed set of available roof types, and its parameters are estimated using the DSM.…"

"(Brenner,…) extracts planar faces from a regularized DSM using a random sampling consensus (RANSAC) approach. Faces are accepted or rejected based on a set of rules… The final topology of the roof is obtained from all accepted regions by a global search procedure."

and on parameter economy:

"…a simple box in 3D space can be described by its position (3), orientation (3) and dimensions (3), for a total of 9 parameters…"

Source: https://www.ikg.uni-hannover.de/fileadmin/ikg/Forschung/publications/Brenner_isprs2004.pdf

What this reaches: (i) a database/library of parametric roof primitives ("a fixed set of available roof types") from which an optimal primitive is selected — claim 1 step 7 and claim 7's "searching a parametric shapes database"; (ii) parameter estimation against a surface model (DSM) — claim 1 step 8/claim 4's fitting; (iii) RANSAC plane extraction from a height/DSM surface; and (iv) topology derivation from accepted planar regions — bridging to 2.3.

2.5 Bernardini et al. 1999 — The Ball-Pivoting Algorithm for Surface Reconstruction

Cited as: IEEE Transactions on Visualization and Computer Graphics, vol. 5, no. 4, Oct.–Dec. 1999, pp. 349–359 (per the patent's own NPL list). BPA is the canonical algorithm for reconstructing a triangle mesh from an unorganized oriented point set by rolling a ball of radius ρ across the points and pivoting about existing edges to form triangles, with boundaries emerging where no further point can be reached.

Confidence note (see § 7): The citation itself is confirmed from the patent's NPL list; the algorithmic characterization is from my training knowledge and my live re-verification query returned no results before the search budget was exhausted. Treat the algorithm description as high-confidence general knowledge, not as a freshly re-fetched quotation.

2.6 Pauly, Gross & Kobbelt 2002 — Efficient Simplification of Point-Sampled Surfaces

Cited as IEEE Visualization, Oct. 27–Nov. 1, 2002, pp. 163–170. This line of work derives local surface approximations from the covariance/moment matrix of a point neighborhood, whose eigenvectors give the local surface direction and normal — the same machinery claim 3 recites. Same confidence caveat as § 2.5.

2.7 The patent-family and platform references (Dimsdale/Cyra; SAIC; Besl; Hayano; Harris)

• Dimsdale/Cyra, US 6,473,079 and US 2004/0051711 / US 2006/0006309 — integrated laser-scanning systems for imaging and modeling 3D objects; the foundational point-cloud acquisition + computer modeling disclosure for claim 1 step 1.
• SAIC, US 2004/0109012 / US 7,398,481 ("Virtual Environment capture") — capture of a physical environment into a virtual (modeled) environment.
• Besl, US 2004/0217956 — processing/compressing/streaming/interactive rendering of 3D color image data (relevant to the "reduce storage vs. LIDAR dataset" motivation).
• Hayano, US 2004/0041805 ("Automatic generating device for 3-D structure shape") — automatic 3D structure-shape generation.
• Harris, US 6,654,690 — automated topographic-model generation from imagery.

Confidence note: These five are mapped from their titles/assignees and the patent's own citation list. I was unable to re-verify their disclosures line-by-line before the search budget was exhausted; the analysis below therefore does not depend on them, and they serve only as corroborating secondary references.

3. Element-by-element mapping of claim 1

Observation: No single reference discloses all eight limitations (compare Brenner's non-LIDAR DSM/ground-plan workflow with Forlani's LIDAR workflow with no primitive library). The case is therefore obviousness, not anticipation — precisely the § 103 posture.

4. Proposed § 103 grounds

Ground A — Claim 1

Rottensteiner 2003 (B1) + Forlani 2003 (B2) + Gruen & Wang 1998 (B3) + Brenner 2004 (B4).

• B1 + B2 together supply limitations 1–5 (point cloud, ground removal, planar-region detection, local planes, roof-structure planes/polygons).
• B3 supplies limitation 6 (topology/face-relation model).
• B4 supplies limitations 7–8 (primitive library selection + parameter fit).

Ground B — Claim 1, alternative

Forlani (B2) + Gruen & Wang (B3) + Brenner (B4) + Dimsdale/Cyra US 6,473,079 (B5). B5 supplies the point-cloud acquisition; B2–B4 supply the rest. Because Forlani alone teaches terrain separation → RANSAC plane segmentation → roof-topology reconstruction → 3D building model, this two-or-three-reference combination is the leanest path to claim 1.

Ground C — Claim 5

Ground A or B + Bernardini et al. (B6). Claim 5's "ball pivoting algorithm in order to define the boundaries" is disclosed by B6, which the examiner itself cited.

Ground D — Claims 3 and 4

Ground A or B + Pauly et al. (B7) for claim 3, and B1/B2 for claim 4:

• Claim 3 (covariance matrix + eigenvalues → local planes): Pauly's covariance/eigen analysis of point neighborhoods; Rottensteiner also computes eigenvalues of local neighborhoods.
• Claim 4 (normal per local plane; group planes with substantially parallel normals): Rottensteiner expressly describes "locally parallel surface normal vectors" and locally planar neighborhoods; Forlani segments roof slopes by gradient orientation. Both are the recited normal-parallel grouping.

Ground E — Claims 2, 6 and 7

• Claim 2 (remove non-planar points; largest-point-count plane = ground): Rottensteiner (surface-roughness/point-like rejection; morphological DTM) + routine 3D connected-components (the specification itself calls this a "3D connected components algorithm").
• Claim 6 (two or more topological graphs): B3 (CC-Modeler processes buildings as roof units, structuring each face model separately) and B4 (complex buildings modeled as combinations of primitives, permitting sub-graph decomposition). The "sub-graph matching" of claim 6's supporting disclosure is the ordinary way to apply a graph-based model to a complex structure.
• Claim 7 (search a parametric shapes database): B4 is a direct hit — "a fixed set of available roof types," from which "an optimal 3D primitive is selected."

5. Motivation to combine (KSR rationales)

The combination is not hindsight; each linkage is supplied by the references and by the field's shared problem statement.

1. Same field, same problem, same data. All of B1–B4 (and B5–B7) sit in aerial-photogrammetry/computer-vision 3D building reconstruction. B1 and B2 are contemporaneous (2003) LIDAR-based building/roof extraction papers; B3 and B4 are building-model-structuring works the examiner cited as pertinent art. KSR forecloses the argument that references from adjacent sub-communities cannot be combined where the field of endeavor and the problem are the same.

2. Design need / market pressure (the patent's own background). The patent concedes manual modeling was "labor intensive," "costly," and performed by "skilled photogrammetrists and graphic artists," and that LIDAR "is increasingly becoming the modality of choice." Every reference responds to that same need: B3 was created to "give the human operator strong computational support in order to generate 3-D city models from aerial images efficiently"; B1/B2 automate LIDAR roof extraction; B4 explicitly evaluates which modeling approach best ensures "correct topology, enforcing geometric regularities and ensuring a given generalization level."

3. Predictable result from combining familiar elements. Substituting LIDAR-derived planar faces (B1/B2) into a face-topology structuring step (B3) yields no more than the expected result — a topologically consistent polyhedral roof — because each reference already performs its step on a comparable point set. KSR, 550 U.S. at 416–17.

4. "Obvious to try" the graph-as-index. Once the roof is expressed as faces + adjacency relations (B3), using that representation as the search key into a library of parametric/prismatic roof types (B4) is one of a finite number of identified, predictable solutions. B4 supplies the library and the selection step; B3 supplies the graph. Combining a known data structure with a known lookup key is the paradigm KSR "obvious to try."

5. Brenner supplies the express motivation for parametric shapes. B4 argues that regularity "is enforced implicitly by the primitives," that a box reduces to 9 parameters, and that CSG/primitive approaches "yield the correct topology." That is the same benefit the patent asserts ("each building can be described by only a few parameters," "a rectangular box shaped building can be described by three parameters"). Where the prior art already articulates the precise advantage the applicant claims, the motivation is express, not inferred.

6. Ball-pivoting is a ready-for-use known technique. B6 is a surface-reconstruction algorithm designed for point clouds. Applying it to define polygon boundaries over planar patches is "use of a known technique to improve a similar device in the same way." KSR, 550 U.S. at 417.

7. Applicant's own admission strengthens the case. The specification states the claimed approach is an alternative to a known RANSAC plane-fitting/edge-alignment method and is preferred only because it "simplifies the rooftop modeling process and mitigates the processing used to align the abutting edges." An improvement framed as simplification/burden-reduction — with no assertion of a new or unexpected technical effect — is exactly the kind of predictable advance § 103 reaches.

6. Anticipated patent-owner rebuttals and how they fare

Where the case for non-obviousness is strongest. If any limitation has genuine patentable weight, it is the graph-keyed retrieval of prismatic shapes (claim 1 steps 6–7 read together, and claim 7). B4 selects primitives from a library, but by ground-plan decomposition and DSM parameter fitting, not by matching a face-adjacency graph. A skilled drafter arguing for the patent would focus there and would point to claim 6's "two or more topological graphs" (sub-graph matching) as the specific, non-trivial algorithmic core. I assess this as a real but modest hurdle: sub-graph matching against a shape database is a standard computer-science technique, and B4's "global search procedure" over accepted regions supplies the analogous step.

7. Bottom line and confidence

Overall assessment. Claims 1–7 of US 7,728,833 are, in my view, more likely than not invalid under § 103 when the NPL on the face of the patent is given its full weight — chiefly because the patent's own admitted background already contains the plane-fitting/RANSAC technique, because Gruen & Wang's CC-Modeler is literally a "topology generator" for the face model of a 3D object, and because Brenner 2004 expressly teaches selecting an optimal parametric roof primitive from a fixed library. The claimed advance reduces largely to substituting a graph-derived selection key and a LIDAR point cloud for the ground-plan/DSM inputs of the cited art — a predictable improvement rather than an unexpected one.

Caveats and confidence limits (stated honestly):

1. Verified vs. unverified content. I live-verified the full text/abstract of B1 (Rottensteiner), B2 (Forlani), B3 (Gruen & Wang), and B4 (Brenner). My search budget was exhausted before I could re-fetch B6 (Bernardini) and B7 (Pauly); their characterizations rest on the patent's own NPL citations plus my training knowledge. The B6/B7 mapping should be re-verified against the papers before being relied on in a filing.
2. Secondary references unverified. Dimsdale/Cyra (B5), SAIC, Besl, Hayano, and Harris are mapped from titles/assignees only. The analysis does not depend on them.
3. Claim construction. My reading of "topology graph" and "parametric shapes" follows the specification and the wording of claims 1 and 7. A narrower construction (e.g., importing the O/S edge-labeling from the specification) would strengthen the patent's position on limitation 6 and weaken Ground A/B accordingly. Conversely, the absence of the labels from the claim text cuts against the patent owner — this is Phillips-style claim-construction risk, not a certainty.
4. Procedural posture. This is a merits analysis only. It does not account for (a) a possible § 101/Alice challenge to claim 1 — a roof-modeling method with only one "by a computer" limitation is an eligibility candidate worth separate treatment — or (b) any IPR/PTAB estoppel or claim-amendment history, which no public record for this patent was found to disclose (see the earlier Litigation section: no suit, IPR, or CAFC appeal located; treat as a negative search result, not a certification).
5. No contradiction with the earlier sections. The reference list and claim breakdown generated previously are consistent with what I found live; the only additions are the substantive disclosures of B1–B4 obtained by re-fetching the sources.

Sources: https://patents.google.com/patent/[US7728833B2](/patent/US7728833B2)/en ; https://www.research-collection.ethz.ch/bitstream/handle/20.500.11850/150916/eth-31183-01.pdf ; https://www.sciencedirect.com/science/article/abs/pii/S0924271698000112 ; https://www.ikg.uni-hannover.de/fileadmin/ikg/Forschung/publications/Brenner_isprs2004.pdf ; https://core.ac.uk/works/6238648/ ; https://core.ac.uk/download/14982344.pdf ; https://www.semanticscholar.org/paper/Building-Reconstruction-and-Visualization-from-Data-Nardinocchi-Forlani/268faca762f20f8197565fb42b9556fda4706dc7 ; https://www.academia.edu/48635719/Building_reconstruction_and_visualization_from_lidar_data