Prosecution Insights
Last updated: April 19, 2026
Application No. 17/746,506

Computer Vision Systems and Methods for Determining Structure Features from Point Cloud Data Using Neural Networks

Non-Final OA §103
Filed
May 17, 2022
Examiner
WANG, YUEHAN
Art Unit
2617
Tech Center
2600 — Communications
Assignee
Insurance Services Office Inc.
OA Round
5 (Non-Final)
83%
Grant Probability
Favorable
5-6
OA Rounds
2y 7m
To Grant
96%
With Interview

Examiner Intelligence

Grants 83% — above average
83%
Career Allow Rate
404 granted / 485 resolved
+21.3% vs TC avg
Moderate +13% lift
Without
With
+12.9%
Interview Lift
resolved cases with interview
Typical timeline
2y 7m
Avg Prosecution
47 currently pending
Career history
532
Total Applications
across all art units

Statute-Specific Performance

§101
4.3%
-35.7% vs TC avg
§103
69.6%
+29.6% vs TC avg
§102
8.3%
-31.7% vs TC avg
§112
6.6%
-33.4% vs TC avg
Black line = Tech Center average estimate • Based on career data from 485 resolved cases

Office Action

§103
DETAILED ACTION Response to Amendment Applicant’s amendments filed on 08 January 2026 have been entered. Claims 1, 11, 14 and 24 have been amended. Claims 12 and 25 has been canceled. Claims 1-11, 13-24 and 26 are still pending in this application, with claims 1 and 14 being independent. Notice of Pre-AIA or AIA Status The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA . Continued Examination Under 37 CFR 1.114 A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 08 January 2026 has been entered. Claim Rejections - 35 USC § 103 In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status. The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action: A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102 of this title, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made. Claim(s) 1-11, 13-24 and 26 is/are rejected under 35 U.S.C. 103 as being unpatentable over US 2019/0377837 A1 to Geomni, Inc. (hereinafter 'Geomni') in view of US 2021/0063578 A1 to NVIDIA Corporation (hereinafter 'NVIDIA') in view of further REJEB et al. (US 20210192254 A1) (hereinafter REJEB) .Regarding claims 1 and 14, Geomni discloses a computer vision system/method for determining features of a structure from point cloud data (generating computerized models of structures using geometry extraction and reconstruction techniques - execute a data fusion process is applied to fuse the raw data, and a geometry extraction process is performed on the fused data to extract features such as walls, floors, ceilings, roof planes, etc - large- and small- scale features are reconstructed by the system into a floor plan (contour) - process exterior features of the structure to automatically identify condition and areas of roof damage, Abstract, para [0002], system obtains raw data scanned by a sensor in communication with the smart phone, such as a series of photos, RGB image data (still, fisheye, panoramic, video, etc.), infrared (IR) image data, mobile sensor data gyroscope/accelerometer/barometer, etc.), laser range data (point cloud data), LIDAR, global positioning system (GPS) data, X-ray data, magnetic field data, depth maps, and other types of data, para [0006], generating three dimensional computer models of structures using geometry extraction and reconstruction techniques - "structure" refers to physical structures such as homes, dwellings, buildings, etc., para [0021)-[0028], Open Computer Vision library (OpenCV), para [0039]-[0040]), comprising: a database storing point cloud data (para [0021]-[0028], system stores the reconstructed structure features in a feature database 86, para [0039]-[0040]); and a processor in communication with the database, the processor programmed to perform the steps of (para [0022]-[0026]): retrieving the point cloud data from the database (system stores the reconstructed structure features in a feature database 86, for subsequent retrieval and usage, para [0039]-[0040]); receiving a geospatial region of interest (, para [0039]-[0040] The information can also be requested by specifying an individual property via an address, a geocode, etc.); selecting a property parcel within the geospatial region of interest (system stores the reconstructed structure features in a feature database 86, for subsequent retrieval and usage, para [0039]-[0040], user captures the entire room in a 3D point cloud - Once the 3D models and all associated data have been extracted, this information can be made available through a database maintained by the system. The information can also be requested by specifying an individual property via an address, a geocode, etc. The information can also be aggregated and reports generated on multiple structures - system could be queried to display a list of all properties in postal code 84097 that have more than 4,000 square feet of living space and a pool. Properties in the database can be viewed online or on any mobile or desktop device, para [0046]-[0052]); processing the point cloud data using a neural network applied over an area of the property parcel to extract a structure or a feature of a structure situated in the area from the point cloud data (para [0021]-[0024], extract structure facade and rooms using concave and convex hull algorithms, identify geometries as real objects using neural networks and then, refine polyhedral geometry, etc, para [0039]-[0040], Identification of the following small-scale features can be accomplished by training and utilizing neural networks (e.g., deep convolutional neural networks), para [0051]-[0055]); and determining at least one attribute of the extracted structure or the feature of the structure using the neural network (The large- and small-scale features are reconstructed by the system into a floor plan (contour) and/or a polyhedron corresponding to the structure – the system can also process exterior features such as roof and wall image data to automatically identify condition and areas of roof damage, para [0006], generating three-dimensional computer models of structures using geometry extraction and reconstruction techniques - "structure" refers to physical structures such as homes, dwellings, buildings, etc., para [0021]-[0024]) but fails to specifically disclose computer vision system/method using point cloud data. However, NVIDIA in analogous art discloses computer vision system/method using point cloud data (Abstract, performing accurate and reliable LiDAR range image deep neural network (DNN) based processing in the form of a combined point cloud segmentation and bounding box regression network ("PCSNet"), para [0004]-[0005], DNN(s) may process the LiDAR data to compute outputs corresponding to instance segmentation masks, per-class semantic segmentations masks, and/or bounding shapes (e.g., two-dimensional (2D) range image bounding boxes). These outputs may be processed into 2D bounding boxes (e.g., corresponding to the LiDAR range image) and/or three-dimensional (3D) bounding boxes (e.g., corresponding to a LiDAR point cloud used to generate the LiDAR range· image) and class labels for the detected objects, para [0022]-[0029], PVA is used to perform computer stereo vision - Many applications for Level 3-5 autonomous driving require motion estimation/stereo matching on-the-fly (e.g., structure from motion, pedestrian recognition, lane detection, etc.). The PVA may perform computer stereo vision function on inputs from two monocular cameras, para [0138]-[0140]). It would have been obvious to one of ordinary skill in the art to add NVIDIA's real-time object detection algorithms to Geomni's computer vision system/method so that by providing a CPU complex, GPU complex, and a hardware acceleration cluster, the technology described herein allows for multiple neural networks to be performed simultaneously and/or sequentially, and for the results to be combined together to enable Level 3-5 autonomous driving functionality (see para [0154]-[0156]). REJEB in analogous art discloses a method includes teaching, based on the dataset, a neural network configured for segmenting an input 3D point cloud including an object (Abstract). REJEB teaches classifying every point of the point cloud data to estimate a probability for each point indicating whether the point belongs to one or more classes (“for each point, the neural network outputs a classification of the point as part of the object or not, together with a confidence score… classifying each point of a point cloud as “object” or “background”, as previously discussed” see para [0180]; “the neural network is train to detect that, the higher the coordinate is, the greater is the probability of the point to belong to the object to select” see para [0148]). REJEB in view of Geomni further teaches segmenting the point cloud data into points indicative of a background (“each point of each 3D point cloud of the training dataset may comprise a single annotation as “object” (i.e. the point belongs to an object, for example said same object) or “background” (i.e. the point belongs to the background), i.e. and no further annotation, e.g. indicating an object category (e.g. such as a chair) for the point. Thus, the neural network is agnostic to the type of the objects in the input 3D point cloud when segmenting it.” see para [0047] of REJEB), a ground field (“In step 154, the system constructs a floor contour and/or polyhedron representing the structure of the room. This is illustrated in FIG. 9D, wherein the extracted walls 166 and floor plane (or, contour) 168 is “assembled” by the system” see para [0049] of Geomni), and a roof structure (“[0226] A second implementation is illustrated by FIGS. 8 and 9. As shown in FIG. 8, a 3D point cloud 80 comprising an object 82 is displayed to a user” see para [0226] of REJEB); It would have been obvious to one of ordinary skill in the art to add REJEB’s classification module to Geomni's computer vision system/method so that be able to classify one or more points of point cloud data into one or more wall points and one or more non-wall points. The point cloud data includes a depiction of an object and background (see para [0047]). By using features of both the classifying and segmentation the point clouds, the system delivers an open object model linking products, processes, resources to enable dynamic, knowledge-based product creation and decision support that drives optimized product definition, manufacturing preparation, production and service (see para [0003]). Regarding claims 2 and 15, Geomni discloses a computer vision system/method wherein the database stores one or more of LiDAR data, a digital image, a digital image dataset, a ground image, an aerial image, a satellite image, an image of a residential building, or an image of a commercial building (Abstract, para [0002], para [0006], system obtains raw data scanned by a sensor in communication with the smart phone, such as a series of photos, RGB image data (still, fisheye, panoramic, video, etc.), infrared (IR) image data, mobile sensor data (gyroscope/accelerometer/barometer, etc.), laser range data (point cloud data), LIDAR, global positioning system (GPS) data, X-ray data, magnetic field data, depth maps, and other types of data, para [0006], para [0021 ]-[0028], system stores the reconstructed structure features in a feature database 86, para [0039]-[0040]). Regarding claims 3 and 16, Geomni discloses a computer vision system/method wherein the processor generates one or more three dimensional representations of the structure or the feature of the structure based on the digital image or the digital image dataset (obtains raw data scanned by a sensor in communication with the smart phone, such as a series of photos - laser· range data (point cloud data), LIDAR, global positioning system (GPS) data, para [0006], generating three-dimensional computer models of structures using geometry extraction and reconstruction techniques - "structure" refers to physical structures such as homes, dwellings, buildings, etc., para [0021]- [0028]). Regarding claims 4 and 17, Geomni discloses a computer vision system/method wherein the structure or the feature of the structure comprises one or more of-a structure wall face, a roof structure face, a segment, an edge, a vertex, a wireframe model, or a mesh model ("structure" refers to physical structures such as homes, dwellings, buildings, etc., para [0021]-[0028], extract only the portions of the imagery that are roof and wall faces, or attempt to to segment the images using roof/wall material detection algorithms or known "grabcut" algorithms. These imagery clips could be sent through neural networks/algorithms to identify possible areas of wind and hail damage and can return a list of coordinates tied to the imagery where the damage is suspected, para [0053]-[0055]). Regarding claims 5 and 18, NVIDIA discloses a computer vision system/method wherein the processor estimates probabilities that the point cloud data belongs to one or more classes to determine if the point cloud data includes the structure, to determine if the structure is damaged, to classify a type of the structure, or to classify one or more objects associated with the structure (DLA may be used to run any type of network to enhance control and driving safety, including for example, a neural network that outputs a measure of confidence for each object detection. Such a confidence value may be interpreted as a probability, or as providing a relative "weight" of each detection compared to other detections. This confidence value enables the system to make further decisions regarding which detections should be considered as true positive detections rather than false positive detections, para [0138]-[0140]). Regarding claims 6 and 19, NVIDIA discloses a computer vision system/method wherein the processor performs semantic segmentation to estimate a probability that a point of the point could data belongs to a class or an object (DNN(s) may process the LiDAR data to compute outputs corresponding to instance segmentation masks, per-class semantic segmentations masks, and/or bounding shapes (e.g., two dimensional (2D) range image bounding boxes), para [0024]-[0029], neural network that outputs a measure of confidence for each object detection. Such a confidence value may be interpreted as a probability, or as providing a relative "weight" of each detection compared to other detections, para [0138]-[0140]). Regarding claims 7 and 20, NVIDIA discloses a computer vision system/method wherein the processor performs instance segmentation to estimate if a point of the point could data belongs to a feature of a structure (DNN(s) may process the LiDAR data to compute outputs corresponding to instance segmentation masks, per-class semantic segmentations masks, and/or bounding shapes (e.g., two-dimensional (2D) range image bounding boxes), para [0024]-[0029], neural network that outputs a measure of confidence for each object detection. Such a confidence value may be interpreted as a probability, or as providing a relative "weight" of each detection compared to other detections, para (0138]-[0140]). Regarding claims 8 and 21, Geomni discloses a computer vision system/method wherein the processor performs a regression task to estimate values of each point of the point cloud data or to estimate roof structure features from the point cloud data (system prepares the point cloud data to extract point cloud normals, to remove outliers, and to remove clearly non-wall points - values correspond with the number of points from the cloud which project onto the cell in the horizontal grid (e.g., flattening the point cloud). Lines are created on the grid based on consecutive cells with point population. These lines are weighted based on cell point population and intersections with neighbor lines, etc. As shown in FIG. 14D, it can be seen that these lines are filtered and adjusted to better fit the real world, displacing them a bit out of the grid, para [0056]-[0060]). Regarding claims 9 and 22, Geomni discloses a computer vision system/method wherein the processor performs an optimization task to improve the point cloud data (Data extraction depends on the ability to identify specific geometric elements in the data set. Specific algorithms with optimal pipeline and parameters could be utilized depending on the nature of input and estimated input accuracy, para [0029]-[0039], region is initialized with the seed from the point cloud and its nearest neighbor in the point cloud 160. At this stage, the region is two points, and from this information, the system calculates the center and normal for an optimal plane of the region, para [0047]-[0052]). Regarding claims 10 and 23, Geomni discloses a computer vision system/method wherein processor improves the point cloud data by increasing a density or resolution of the point cloud data, providing missing point cloud data, and filtering noise (system prepares the point cloud data to extract point cloud normals, to remove outliers, and to remove clearly non-wall points - values correspond with the number of points from the cloud which project onto the cell in the horizontal grid (e.g., flattening the point cloud). Lines are created on the grid based on consecutive cells with point population. These lines are weighted based on cell point population and intersections with neighbor lines, etc. As shown in FIG. 14D, it can be seen that these lines are filtered and adjusted to better fit the real world, displacing them a bit out of the grid, para [0056]-[0060]). Regarding claims 11 and 24, Geomni discloses a computer vision system/method wherein the geospatial region of interest (ROI) is specified by a user (system stores the reconstructed structure features in a feature database 86, for subsequent retrieval and usage, para [0039]-[0040], user captures the entire room in a 3D point cloud – Once the 3D models and all associated data have been extracted, this information can be made available through a database maintained by the system. The information can also be requested by specifying an individual property via an address, a geocode, etc. The information can also be aggregated and reports generated on multiple structures, para [0046]-[0052]). Regarding claims 13 and 26, Geomni discloses a computer vision system/method wherein the processor preprocesses the point cloud data by performing one or more of: spatially cropping the point cloud data, spatially transforming the point cloud data, down sampling the point cloud data, removing redundant points from the point could data, up sampling the point cloud data, filtering the point cloud data, projecting the point cloud data onto an image to obtain a two-dimensional representation, obtaining a voxel grid representation, or generating a new feature from the point cloud data (system prepares the point cloud data to extract point cloud normals to remove outliers, and to remove clearly non-wall points - values correspond with the number of points from the cloud which project onto the cell in the horizontal grid (e.g., flattening the point cloud). Lines are created on the grid based on consecutive cells with point population. These lines are weighted based on cell point population and intersections with neighbor lines, etc. As shown in FIG. 14D, it can be seen that these lines are filtered and adjusted to better fit the real world, displacing them a bit out of the grid, para [0056]-[0060]). Response to Arguments Applicant's arguments filed on 08 January 2026, with respect to the 103 rejection have been fully considered but they are not persuasive. On page 4, Applicant's Remarks, with respect to claims 1 and 14, the applicant argues Lewis et al., the primary reference, fails to teach or suggest receiving a geospatial region of interest, selecting a property parcel within the geospatial region of interest, and processing the point cloud data using a neural network applied over an area of the property parcel to extract a structure or a feature of a structure situated in the area from the point cloud data, as recited by amended independent Claims 1 and 14. The examiner respectfully disagrees with this argument. In Lewis, the information can also be requested by specifying an individual property via an address, a geocode, etc, that equals the function of receiving a geospatial of ROI. Further, Lewis disclosed the user captures of an entire room in a 3D point cloud. The captures activity reads on the limitation of obtains point cloud data of a structure. REJEB further disclosed Learning a neural network configured for segmenting an input 3D point cloud comparing an object, which means the point clouds covers an area that larger than a specific object, and the object is situated within the range of the point cloud. Regarding this argument, it is respectfully noted that the combination of Lewis and REJECB teaches all the limitations of claimed, as amended. Conclusion Any inquiry concerning this communication or earlier communications from the examiner should be directed to Samantha (Yuehan) Wang whose telephone number is (571)270-5011. The examiner can normally be reached Monday-Friday, 8am-5pm. Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice. If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, King Poon can be reached on (571)272-7440. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300. Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000. /Samantha (YUEHAN) WANG/ Primary Examiner Art Unit 2617
Read full office action

Prosecution Timeline

May 17, 2022
Application Filed
Aug 01, 2022
Response after Non-Final Action
Sep 26, 2023
Non-Final Rejection — §103
Mar 27, 2024
Response Filed
Apr 24, 2024
Final Rejection — §103
Oct 25, 2024
Request for Continued Examination
Oct 29, 2024
Response after Non-Final Action
Dec 09, 2024
Non-Final Rejection — §103
Jun 10, 2025
Response Filed
Jul 03, 2025
Final Rejection — §103
Jan 08, 2026
Request for Continued Examination
Jan 21, 2026
Response after Non-Final Action
Mar 10, 2026
Non-Final Rejection — §103 (current)

Precedent Cases

Applications granted by this same examiner with similar technology

Patent 12597178
VECTOR OBJECT PATH SEGMENT EDITING
2y 5m to grant Granted Apr 07, 2026
Patent 12597506
ENDOSCOPIC EXAMINATION SUPPORT APPARATUS, ENDOSCOPIC EXAMINATION SUPPORT METHOD, AND RECORDING MEDIUM
2y 5m to grant Granted Apr 07, 2026
Patent 12586286
DIFFERENTIABLE REAL-TIME RADIANCE FIELD RENDERING FOR LARGE SCALE VIEW SYNTHESIS
2y 5m to grant Granted Mar 24, 2026
Patent 12586261
IMAGE PROCESSING APPARATUS, IMAGE PROCESSING METHOD, AND STORAGE MEDIUM
2y 5m to grant Granted Mar 24, 2026
Patent 12567182
USING AUGMENTED REALITY TO VISUALIZE OPTIMAL WATER SENSOR PLACEMENT
2y 5m to grant Granted Mar 03, 2026
Study what changed to get past this examiner. Based on 5 most recent grants.

AI Strategy Recommendation

Get an AI-powered prosecution strategy using examiner precedents, rejection analysis, and claim mapping.
Powered by AI — typically takes 5-10 seconds

Prosecution Projections

5-6
Expected OA Rounds
83%
Grant Probability
96%
With Interview (+12.9%)
2y 7m
Median Time to Grant
High
PTA Risk
Based on 485 resolved cases by this examiner. Grant probability derived from career allow rate.

Sign in with your work email

Enter your email to receive a magic link. No password needed.

Personal email addresses (Gmail, Yahoo, etc.) are not accepted.

Free tier: 3 strategy analyses per month