DETAILED ACTION
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 .
Claim status
Claims 1-2 and 5-22 are pending. Claims 1 -2 and 5 -13 and 15 and 17-20 were amended. Claims 14 and 16 remain in their original form. Claims 3 and 4 were canceled. Claims 21 and 22 are new.
Response to Arguments
Specification and Drawing
Applicant’s amendments to the Specification and Drawings have overcome each and every objection previously set forth in the Non-Final Office Action mailed 9 April 2026.
35 USC § 101
Claims 1, 11 and 20 have been analyzed for patent eligibility and found to recite a judicial exception but are integrated into a practical application as amended. The rejection of 9 April 2026 has been withdrawn.
The analysis concluded that under step 2A Prong Two of the Subject Matter Eligibility Test, claim 1, 11 and 20 integrate the recited judicial exception into a practical application by reciting performing a drilling operation based on the digitization of the curve mask generated by the trained digitization neural network.
Therefore, claims 1, 11 and 20 and all the dependent claims are eligible at Prong Two of Step 2A (see MPEP 2106.04(d))).
35 USC § 103
Applicant’s amendments with respect to claims 1, 11 and 20 have been fully considered. Applicant argues that the cited references fail to teach the claimed two-stage neural network architecture, predictive-row wise pixel indexing, and scaling based on well log header information (see page 18). Arguments regarding two-stage neural network architecture and scaling based on well log header information are not persuasive as Katole in view of Yin teaches these limitations. Examiner agrees that Katole in view of Yin does not teach row-wise pixel index generation, as required by the amended claims. The rejection of 9 April 2026 under 35 USC 102 has been withdrawn.
However, further search finds teaching of this approach, as discussed below.
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 (i.e., changing from AIA to pre-AIA ) 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, 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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or non-obviousness.
Claims 1-2, 5-9, 11-18 and 20-21 are rejected under 35 U.S.C. 103 as being unpatentable over Atul Laxman Katole et al. (WO2021178412) hereinafter Katole in view of Kun Yan Yin et al. (US20220010675) hereinafter Yin and M. Quamer Nasim (VeerNet: Using Deep Neural Networks for Curve Classification and Digitization of Raster Well-Log Images, J. Imaging 2023, Vol 9, Issue 7, 136) hereinafter Nasim.
Regarding claim 1, A machine learning method of digitizing a well log curve (method for digitizing a raster image¶ [10], well log¶ [80], curves ¶ [6] and fig. 4A).
Katole teaches utilizing a machine learning model for its segmentation and digitization steps, and furthermore characterizes its machine learning model as deep learning module,¶ [18].
However, Katole does not disclose that the machine learning model is a neural network.
Yin teaches that image analysis-based well log data generation (fig.2 and ¶ [8]), including
feature identification and translation of image data into numerical values (fig. 2 and ¶ [34]), may be
implemented using neural networks (fig. 2)
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to implement the machine learning model of Katole using neural networks as taught by Yin because neural networks improve the accuracy of the curve identification and enable reliable translation of image-based representations into numerical data.
Katole in view of Yin further teaches the machine learning method comprising: obtaining a scan of a well log curve paper document, the well log curve paper document including the well log curve; (receiving an image file including one or more target object (Katole, ¶ [4]). The image file includes a raster image including pixels, and the one or more target objects include one or more curves that represent measurements of a subsurface property. (Katole, ¶ [6]) )
Katole in view of Yin teaches providing the scan of the well log curve paper document
(Katole, blocks 804 and 806, fig. 8) to a trained segmentation neural network (Katole, object segmentation machine learning model ((model 530 and module 426 and 428) and ¶ [4-6])).
Katole in view of Yin further teaches generating (the output of the first stage in Katole’s pipeline is a Raster image with only curves, where the background has been removed, leaving only the pixels corresponding to the isolated curves, Katole, fig. 5 (intermediate raster image (plot lines))), via the trained segmentation neural network (denoising model 525, Katole ¶ [67] and the model 525 is a trained model, ¶ [78]), a curve mask from the scan by isolating the well log curve; ( Generating the intermediate image comprises removing pixels that represent the noise, Katole, ¶ [6]); wherein a curve mask is obtained (block 406 fig.4A, extracted curves fig.5 and ¶ [69]).
Katole teaches the distinct steps of segmentation and digitization as a sequence within its machine learning workflow.
However, Katole does not explicitly discloses utilizing distinct machine learning models for
segmentation and digitization.
Yin further teaches that its system 200 utilizes multiple distinct neural networks (fig2. Block 270
and ¶ [55]) to perform separate functions where the task of segmentation (identifying and isolating a
curve) is handled by the image analysis engine 250 (fig. 2) and digitization is handled by the image
decoder 280 (fig. 2) in conjunction with neural network 270 that is trained based on a set of training
data 230 (fig. 2).
It would have been obvious to a person having ordinary skill in the art before the
effective filing date of the claimed invention to implement the machine learning model of Katole for
segmentation and digitizing using distinct neural networks as taught by Yin because utilizing distinct
network allows each model to be specialized for its specific goal, rather than using a single generic
model that may be less efficient at both.
Katole in view of Yin therefore teaches providing (Katole, block 808 provides its output directly to block 810, fig 8) the curve mask (Katole, extracted curves fig. 5) to a trained (¶ [16] and ¶ [59]) digitization neural network (Katole, pixel to unit convertor 446 fig. 4B or the object extraction (machine learning) model 530 fig. 5 performing discretization).
Katole in view of Yin teaches generating, via the trained digitization neural network (Katole utilizes a trained machine learning model within a raster digitization module (908) to process isolated curves,¶ [16, 91]), a digitization of the curve mask ( describes processing the extracted individual curves (the pixels isolated by the segmentation model) to obtain digitized data stored in a file, Katole, ¶ [63]), wherein generating the digitization (transform well log or other image-based data (e.g., plots and associated metadata) from raster images into digital data, Katole ¶ [18]) of the curve mask (Katole feeds an intermediate raster image (plot lines) into final digitization model, fig. 5).
Katole in view of Yin further teaches using a model to identify curve ¶ [6] pixels or discrete elements from the isolated curve, (Katole ¶ [62 & 63]) ((the digitization of the curve mask) includes the respective pixel). Katole in view of Yin utilizes a specific module called pixel to unit converter 446. This module is designed to take the extracted pixel locations and map them to plot values,(Katole, ¶ [63]) (scaling). Katole in view of Yin teaches that the digitization step uses metadata extracted from the log header (including minimum/maximum ranges and scales) ,(Katole, ¶ [73]) (based on a scale and range of the well log curve as indicated in a well log header). Katole in view of Yin describes transforming the extracted curves pixels to numerical, (Katole, ¶ [82]) (to determine a numerical value representing the well log curve).
However, Katole in view of Yin does not teach a model that treats each image row as an input to be classified into exactly one class, where that class is the categorial pixel index. Katole in view of Yin does not teach generating, for each row of pixels from the curve mask, a respective pixel index.
Nasim teaches a digitization model that uses Sparse Cross Entropy (SCE) loss (table 2, page 6). SCE loss used specifically when a model is trained to predict a single integer class index (the category) out of N possible choices. For well-log digitization, the categories are the horizontal pixel columns. Therefore, by using SCE loss, Nasim can predict a respective pixel index for every interval of the independent variable (the row). As an example, see fig. 4 and 5.
Furthermore, it describes multi-class segmentation where each mask contains classes (e.g., 0, 1, 2) corresponding to the background and individual curves (page 7, ¶ [3]), predicting which class (pixel index ) a curve belongs to in specific row. (generating, for each row of pixels from the curve mask, a respective pixel index)
Nasim further teaches extracting a 1D signal from generated spatial masks (page 4, ¶ [4]) (the respective pixel index). In the context of well-log digitization, this 1D signal represents the horizontal spatial coordinates (the pixel index) of the curve for every depth level.
Nasim explains that the image in fig. 1(b) is used to identify the specific information required for scaling, such as the variable being measured (e.g., Gamma Ray) and its associated numerical scale (scaling). Figure 1(c) additionally shows the depth line coding, which provides the reference for the vertical axis (the depth range) used in the digitization process.
Nasim further teaches that VeerNet workflow as a simple four step process, which includes verifying the scales and range values of curves immediately before saving the digitized values (page 11, ¶ [8]).( to determine a numerical value representing the well log curve)
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to modify Katole’s hierarchical pipeline with Nasim’s classification methodology because using transformers and attention mechanisms, as taught by Nasim, is a solution for high-resolution imagery to balance retaining key signals while reducing dimensionality, leading to more robust digitization.
Katole in view of Yin further teaches wherein a digitization of the curve mask is obtained (Katole, block 408, fig. 4A); and outputting the digitization of the curve mask (Katole, block 448 fig. 4B or block 812 fig. 8).
Katole further teaches that the legacy well log data (output of their machine learning model which is a numerical value) may be employed for planning well locations and trajectory and equipment (¶ [86]). Katole also teaches a communication unit 134 (¶ [26]) that communicates with drilling tool to send command and to receive data. The surface unit 134 may collect and use data in real- time (¶ [31]) and it may also be provided with one or more controllers to actuate mechanism in oilfield (¶ [32]). (performing a drilling operation )
However, Katole does not explicitly teach performing a drilling operation based on the digitization of the curve mask generated by the trained digitization neural network.
Yin teaches that geologist and miners require data in real-time to make a determination on drilling ¶ [21]. It further specifies that the numerical output is used by a geological entity for performing one or more mining operations ¶ [23].( based on the digitization of the curve mask generated by the trained digitization neural network)
It would have been obvious to a person having ordinary skill in the art before effective filing date of the claimed invention to modify the method of Katole in view of Yin and Nasim to perform a drilling operation, as taught by Yin, in order to enable real-time decision-making and improve operational efficiency. Such a modification represents the application of known data-driven control of drilling operations to the system of Katole in view of Yin and Nasim, yielding predictable results.
Claims 11 and 20 recite limitations similar to those in claim 1, but are directed to a computer readable medium and a system, respectively, comprising a processor configured to perform the recited method steps. Katole in view of Yin teaches a computing system including one or more processors and memory configured to perform the digitization method (Katole, fig 5 and fig 9) as well as a non- transitory computer readable medium storing instructions executable by a processor to perform the same operations (Katole, storage media 906, fig 9 and ¶ [90]). Inclusion of a processor and memory to execute the method steps represents no more than the implementation of the method on generic computer system, which would have been obvious to a person of ordinary skill in the art. Therefore, claims 11 and 20 are rejected for the same reasons as claim 1.
Claim 2 is rejected for the same reasons as set forth with respect to rejection of claim 1.
Claim 12 recites the same limitations as claim 2 in computer-readable form and is rejected for the same reasons discussed above.
Regarding claim 21, Katole in view of Yin and Nasim teaches the method of claim 1 as set forth with respect to rejection of claim 1.
Katole in view of Yin and Nasim teaches a learning portion 422, where machine learning models are trained to eventually map discrete portions of the extracted curves to plot values (Katole, fig. 4B and ¶ [62]) (wherein the trained digitization neural network is trained). The combination treats curve identification and separation as a multi-class classification problem (Nasim, page 7, section 5). To optimize this classification, it uses Sparse Cross Entropy (SCE) loss (Nasim, table 2)(determining, for each row of the pixels from the curve mask a classification)(SCE loss is used to train a model to predict an integer class index (the classification) for an input, which corresponds to the horizontal position for a given row). It further teaches that once the signals are extracted via classification mask, it extract a 1D signal and saves it in CSV/LAS files (Nasim, page 4, ¶ 4) (and generating, for each row of the pixels from the curve mask, a numerical value corresponding to the well log curve based on the classification.). These files contain the final digitized values of the well-log curve (Nasim, page 4, ¶ 5) (a numerical value)
Regarding claim 5, Katole in view of Yin and Nasim teaches the machine learning method of claim 1.
Katole in view of Yin and Nasim teaches a Raster segmentation 426 and 428 (fig. 4)(the trained segmentation neural network). It discloses training machine learning models using training corpus (Katole, block 427 and 429, fig. 4B) and it specifies receiving training raster images (Katole, ¶ [75]) and curve labels (Katole, ¶ [77]) for training to its digitization system 520 where digitization refers to mapping discrete coordinates to pixels (Katole, ¶ [72]).
Katole does not teach using synthetic data for training of its machine learning method. Katole does not teach the trained segmentation neural network is trained using synthetic training data generated from numerical coordinate data.
Yin further teaches training the neural network based on images (¶ [54]) and it specifies that the system use training data 230 (fig. 2) to adjust weights and biases via back propagation. Yin further teaches that neural network 270 may perform a visual reconstruction of the well log to generate training data without presence of any historical data (¶ [36]) (trained using synthetic training data). Yin also teaches that the log curve generator 240
takes dimensions in numerical format and transforms them into images to create training dataset (¶ [35]) (generated from numerical coordinate data.).
It would have been obvious to a person having ordinary skill in the art before effective filing date of the claimed invention to incorporate the synthetic data generation techniques of Yin into the training process of Katole in view of Yin and Nasim in order to augment training databases, thereby improving the robustness and performance of the machine learning models.
Katole in view of Yin and Nasim further teaches wherein each synthetic training datum comprises a respective pair (Katole, pairs of raster images and labels ¶ [64]), wherein each respective pair comprises an image of a respective curve (Katole teaches utilizing pairs of raster images and labels for training of both machine learning models 426 and 428 where the training process involves receiving a training raster image ¶ [75] (the curve with grid noise) and an intermediate label with ground truth (the clean curve) ¶ [76] to teach the model to remove background artifacts.)
Katole in view of Yin and Nasim does not teach generated from respective numerical coordinate data and the image of the respective curve generated from the respective numerical coordinate data combined with a respective grid depiction, and wherein the trained segmentation neural network is trained to remove a grid from the scan of the well log curve paper document.
Yin further teaches a log curve generator 240 that is configured for generating visual curves from mathematical values ¶ [35] to create training set ¶ [36] and inputting the one or more dimensions (dimensions are in numerical format) into an image ¶ [35].
It would have been obvious to a person having ordinary skill in the art before effective filing date of the claimed invention to combine the training framework of Katole in view of Yin and Nasim with the synthetic data generation of Yin to automatically produce the clean labels required by Katole training pipeline from existing numerical coordinate data because manual labeling (tracing curves) is time intensive.
Therefore, Katole in view of Yin and Nasim teaches trained using synthetic training data generated from respective numerical coordinate data (Yin, ¶ [35]) and the image of the respective curve (an intermediate label with ground truth (the clean curve) ¶ [76]) generated from the respective numerical coordinate data (Yin, ¶ [35]) combined with a respective grid depiction (Katole, training raster image ¶ [75], provided as the input half of the pair includes noise such as grid lines. It also notes that the ground truth of the curve location and the noise are known a priori ¶ [76]), and wherein the trained segmentation neural network is trained to remove a grid from the scan (Katole, denoising model 925, 1 [70]) of the well log curve paper document.
Claim 14 recites the same limitations as claim 5 in form of computer readable medium. Claim 14 is rejected for the same reasons set forth with respect to claim 5.
Regarding claim 6, Katole in view of Yin and Nasim teaches the machine learning method of claim 1.
Katole in view of Yin and Nasim discloses training machine learning models using training corpus (Katole, block 427 and 429, fig. 4B) and it specifies receiving training raster images (Katole, ¶ [75]) and curve labels (Katole, ¶ [77]) for training to its digitization system 520 where digitization refers to mapping discrete coordinates to pixels (Katole, ¶ [72]).
Katole in view of Yin and Nasim does not teach using synthetic data for training of its machine learning method. It does not teach the trained digitization neural network is trained using synthetic training data generated from numerical coordinate data.
Yin further teaches training the neural network based on images (¶ [54]) and it specifies that the system use training data 230 (fig. 2) to adjust weights and biases via back propagation. Yin further teaches that neural network 270 may perform a visual reconstruction of the well log to generate training data without presence of any historical data (¶ [36]) (trained using synthetic training data). Yin also teaches that the log curve generator 240
takes dimensions in numerical format and transforms them into images to create training dataset (¶ [35]) (generated from numerical coordinate data.).
It would have been obvious to a person having ordinary skill in the art before effective filing date of the claimed invention to incorporate the synthetic data generation techniques of Yin into the training process of Katole in view of Yin and Nasim in order to augment training databases, thereby improving the robustness and performance of the machine learning models.
Katole in view of Yin and Nasim further teaches that the trained digitization neural network (Katole, pixel to unit convertor 446 fig. 4B or the object extraction (machine learning) model 530 fig. 5 performing discretization) is trained using the synthetic training data generated from numerical coordinate data(Yin, ¶ [35, 36 & 54]), wherein each synthetic training datum of the synthetic training data comprises a respective pair (Katole in view of Yin and Nasim teaches utilizing pairs of raster images and labels for training. It describes a learning portion 422 where the system is fed a training corpus (Katole, ¶ [60])).
Katole teaches utilizing pairs of training "intermediate raster image" (block 730, fig 7) that consist of extracted curve pixels ¶ [62 and 72] and "curve labels" (block 740, fig 7) as ground truth. The system extract curve segments as intermediate input and determines plot values for discrete points along a one or more curves ¶ [6 and 63].
Katole in view of Yin and Nasim does not teach and wherein each respective pair comprises a section of an image of a respective curve generated from respective numerical coordinate data and a corresponding respective numerical position value.
Yin teaches that the network is provided with a specific window (a 128 pixel by 250 pixel image input that contains a curve) ¶ [51] and compares its output against a desired output y(x) which represents the exact numerical position value for that section of the curve ¶ [52]. The log curve generator 240 generates the images for training from mathematical values ¶ [35].
It would have been obvious to a person having ordinary skill in the art before effective filing date of the claimed invention to implement the structure taught by Yin in Katole in view of Yin and Nasim training process to improve accuracy.
Katole in view of Yin and Nasim teaches and wherein each respective pair comprises a section of a respective curve (Yin, ¶ [51]) generated from respective numerical coordinate data (Yin, ¶ [35]) and a corresponding respective numerical position value (Yin, ¶ [52]).
Claim 15 recites the same limitations as claim 6 in form of computer readable medium. Claim 15 is rejected for the same reasons set forth with respect to claim 6.
Claim 7 is rejected for the same reasons set forth with respect to rejection of claim 5 and claim 6.
Regarding claim 8, Katole in view of Yin and Nasim teaches the machine learning method of claim 1, wherein the scan of the well log curve paper document comprises a plurality of different curves (Katole, fig 6A and 6B, two curves 625 and 627), and wherein the trained segmentation neural network (Katole, trained object segmentation model identifies objects (curves) in an intermediate image ¶ [79] and ¶ [83]) is trained to generate the curve mask for a selected line style (Katole, the header segment 624 provides information related to line style 626 and that this metadata is used to distinguish between individual target objects (curves) ¶ [73 & 79]).
It would have been obvious to a person having ordinary skill in the art before effective filing date of the claimed invention to train the segmentation neural network of Katole in view of Yin and Nasim to produce curve masks for a selected line style, as the combination teaches distinguishing between multiple curves in a well log using metadata such as line style. Training the segmentation model to identify and isolate curves based on such distinguishing features represents a predictable implementation of known image segmentation techniques for improving curve differentiation.
Claim 17 recites the same limitations as claim 8 in form of computer readable medium. Claim 17 is rejected for the reasons set forth with respect to claim 8.
Regarding Claim 9, Katole in view of Yin and Nasim teaches the machine learning method of claim 8 as set forth with respect to rejection of claim 8, further comprising annotating the scan of the well log curve paper document with the selected line style (Katole, Metadata 540 which includes information such as line style may be manually added to the system ¶ [77] and also user input for a line style ¶ [73]) prior to the providing the scan of the well log curve paper document to the trained segmentation neural network (Katole, fig 4B illustrates user inputs entering the system as an initial part of the workflow).
It would have been obvious to a person having ordinary skill in the art before effective filing date of the claimed invention to annotate the scan with a selected line style prior to segmentation, as Katole in view of Yin and Nasim teaches incorporating user-provided metadata, including line style, into the processing workflow. Providing such annotations before input to the segmentation model represents a predictable use of known preprocessing techniques to improve feature identification accuracy.
Claim 18 recites the same limitations as claim 9 in form of computer readable medium. Claim 18 is rejected for the reasons set forth with respect to claim 9.
Regarding claim 13, Katole in view of Yin and Nasim teaches the non-transitory computer-readable medium of claim 11 as set forth with respect to rejection of claim 11,
Katole in view of Yin and Nasim discloses training machine learning models using training corpus (block 427 and 429, fig. 4B) and it specifies receiving training raster images (¶ [75]) and curve labels (¶ [77]) for training to its digitization system 520 where digitization refers to mapping discrete coordinates to pixels ¶ [72]. Katole does not teach using synthetic data for training of its machine learning method.
Katole does not teach wherein at least one of the trained segmentation neural network or the trained digitization neural network is trained using synthetic training data, wherein the synthetic training data is generated from numerical coordinate data.
Yin further teaches training the neural network based on images (¶ [54]) and it specifies that the system use training data 230 (fig. 2) to adjust weights and biases via back propagation. Yin further teaches that neural network 270 may perform a visual reconstruction of the well log to generate training data without presence of any historical data (¶ [36])( trained digitization neural network is trained using synthetic training data). Yin also teaches that the log curve generator 240 takes dimensions in numerical format and transforms them into images to create training dataset (¶ [35]) (generated from numerical coordinate data representing well log curves).
It would have been obvious to a person having ordinary skill in the art before effective filing date of the claimed invention to incorporate the synthetic data generation techniques of Yin into the training process of Katole in view of Yin and Nasim in order to augment training databases, thereby improving the robustness and performance of the machine learning models.
Claim 16 have the same limitations as claim 7, in form of computer readable medium. Therefore, it is rejected for the same reasons set forth with respect to claim 7.
Claims 10 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Katole in view of Yin and Nasim and further in view of Ferederick Johnnes Venter et al. (US10095926 B1) hereinafter Venter.
Regarding claim 10, Katole in view of Yin and Nasim teaches the machine learning method of claim 1 and providing the scan of the well log curve paper document to the trained segmentation neural network (for the same reasons discussed with respect to rejection of claim 1)
Katole in view of Yin and Nasim does not teach the tilling architecture.
Katole in view of Yin and Nasim does not teach wherein the providing the scan of the well log curve paper document to the trained segmentation neural network comprises dividing the scan of the well log curve paper document into a plurality of scan tiles and providing the scan tiles individually to the trained segmentation neural network, wherein the curve mask comprises a plurality of curve mask tiles, and wherein the providing the curve mask to the trained digitization neural network comprises providing the curve mask tiles individually to the trained digitization neural network, wherein the digitization of the curve mask comprises a plurality of digitizations of the curve mask tiles.
Venter teaches a hierarchical processing workflow, which utilizes tiles to handle the large scale of well log scans.
Venter teaches dividing the scan of the well log curve paper document into a plurality of scan tiles (Fig. 4, step 404, Col. 5, ll. 33-37) and providing the plurality of scan tiles individually to the neural network (the trace analysis engine 218 (can be implemented as neural network Col. 7, ll. 40-42) can assign tile locations within a graph Col. 7, ll. 53-58), wherein the curve mask comprises a plurality of curve mask tiles (describes processing tiles to produce clusters separated by blank spaces within each tile location, Col. 5, ll. 52-59), and wherein providing the curve mask to the neural network (digitizing contents of tiles by converting a position to a paired value, Col. 2, ll. 1-8 and Col. 8, ll. 6-12) further comprises providing the curve mask tiles individually (clusters assigned to the well log trace based on the probability factor for each tile, Col. 3, ll. 54-67 and Col. 8, ll. 6-12) to the neural network, wherein the digitization of the curve mask comprises a plurality of digitizations of the plurality of curve mask tiles (digitized data can be determined from each of the graphs(tiles) within the well log image and stored collectively in the storage medium).
It would have been obvious to a person having ordinary skill in the art before effective filing date of the claimed invention to incorporate the tiling architecture of Venter into the method of Katole in view of Yin and Nasim in order to efficiently process large well log images and improve computational performance and scalability.
Claim 19 recites the same limitations as claim 10 in form of computer readable medium. Claim 19 is rejected for the reasons set forth with respect to claim 10.
Claim 22 is rejected under 35 U.S.C. 103 as being unpatentable over Katole in view of Yin and Nasim and further in view of Bo Yuan and et al.(Digitization of Well-Logging Parameter Graphs Based on Gridlines-Elimination Approach, Journal of Software, 2019 Vol. 14 (12), page 573-578 ) hereinafter Yuan.
Regarding claim 22, Katole in view of Yin and Nasim teaches the method of claim 1 as set forth with respect to rejection of claim 1.
Katole in view of Yin and Nasim utilizes a pixel to unit converter 446 for scaling (Katole, fig. 4B). It teaches a header segment 624 that includes information configured to establish a scale 628 to permit mapping of values to the plot curves (Katole, ¶ [73]). It extracts a type of data/information represented a minimum and maximum range of plot values from the header to perform the conversion (Katole, ¶ [73]) (scaling the respective pixel index based on the scale and range indicated in the well log header).
However, Katole in view of Yin and Nasim doesn’t explicitly discloses applying a linear transformation based on scale and range.
Yuan provides explicit equations for converting horizontal (l) and vertical (v) pixel positions into physical X and Y coordinates with formula (5) and (6). Formula (5) is a standard linear transformation that scales a raw index based on a minimum and maximum range (smin , smax)( applying a linear transformation based on the scale and range).
It would have been obvious to a person having ordinary skill in the art before effective filing date of the claimed invention to use Yuan’s linear transformation equations within Katole’s in view of Yin and Nasim hierarchical engine because they provide the necessary mathematical solution within muti-class classification problem that predicts a specific class index for every row. This combination results in the fully automates system to replace time-consuming manual tracing.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SAEEDE NAFOOSHE whose telephone number is (571)272-8629. The examiner can normally be reached Monday-Friday 8:00 am -5:00pm.
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, Andrew Schechter can be reached at 571-272-2302. 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.
/SAEEDE NAFOOSHE/Examiner, Art Unit 2857
/ANDREW SCHECHTER/Supervisory Patent Examiner, Art Unit 2857
Prosecution Insights