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 .
Response to Arguments
W/r to the reference of Bian:
Applicant’s arguments, see pages 7-8, filed 3/6/26, with respect to the rejection(s) of claim(s) (1-11) under 35 USC 102 have been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made in view of Totsuka.
Applicant asserts “The Office Action appears to equate Bian's "generating a first prediction" with the "screening step" of claim 1. Applicant respectfully asserts that this conclusion is factually incorrect. In Bian, the "first prediction" is generated by analyzing the first set of images "one at a time as inputs in a forward propagation direction through layers of artificial neurons in an artificial neural network" (See Bian [005l]). As Bian utilizes an artificial neural network for its first determination, it cannot be the high-speed, non-CNN screening mechanism claimed and described in the current application (Spec. [0023]). Furthermore, Bian does not use its first stage to "screen" (filter or reduce) candidates for a second stage. Instead, Bian's separately analyzes a second set of images taken from a different perspective to generate a second prediction image, and these two distinct prediction images are then merged to detect defects (See Bian [0005] and [0051]). The U.S. Court of Appeals for the Federal Circuit has held that the PTO "may not... resort to speculation, unfounded assumptions or hindsight reconstruction to supply deficiencies in its factual basis" (In re Warner). Furthermore, the BRI standard "does not give the PTO an unfettered license to interpret claims to embrace anything remotely related to the claimed invention" (Inre Suitco Surface, Inc.). The Office Action's interpretation that Bian's separate neural network predictions constitute a "screening step" to reduce data for a subsequent CNN is therefore believed to be unreasonable and speculative. As Bian merely performs two defect determinations under different lighting conditions (See Bian Abstract), and fails to disclose the claimed computational screening hierarchy, under proper analysis, Bian does not anticipate independent claims 1, 4, 6, 7, 8, and 10, nor their dependent claims.
The examiner respectfully agrees. However, a new ground of rejection has been issued. (See rejection below)
W/r to claim 9 (101):
Applicant's arguments filed 3/6/26 have been fully considered but they are not persuasive.
Applicant asserts “claim 9 integrate the judicial exception into a practical application”.
The examiner respectfully disagrees. The 101 rejection applied to claim 9 was a different type of 101 rejection. For example, claim 9 recites “A learning model…”. It is claiming a learning model. Specifically, a learning model is not a process because it is not a serial steps, 2) has no physical structure, thus it does not fit within the definition of a machine, 3) is not a matter and therefore is not a composition of matter, and 4) it does not fit the definition of manufacture. As a result, the rejection stands.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claim 9 is rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter. The claim(s) does/do not fall within at least one of the four categories of patent eligible subject matter because claim 9 fails to fall within a statutory category of invention because the claim is directed to software per se. The specs itself does not contain any definition that would exclude “learning model” as software per se.
Claim Rejections - 35 USC § 103
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.
Claim(s) (1-11) are rejected under 35 U.S.C. 103 as being unpatentable over Bian et al (hereinafter Bian)(US Publication 2020/0175669 A1) in view of Takayuki Totsuka et al. (hereinafter Totsuka)(WO 2020/175666 A1 – See Machine translation)
Re claim 1, Bian discloses an inspection method of detecting surface defects on an inspection target (See fig. 9 where it teaches detecting the presence of defects in a work piece.), the inspection method including: an imaging step of acquiring an image of a surface of the inspection target (See fig. 1: 108; fig. 4: 302, 304; fig. 9: 902, 904; ¶ 20, 28, 43, 52; 86 where it teaches imaging device configured to obtain images of a work piece.); an extraction step of extracting defect candidate parts from the image. (See fig. 4; fig. 9: 906, 908; ¶ 55, 58-59, 63 where it teaches generating images 406 which represent defect candidates.)
But the reference of Bian fails to teach a screening step of screening the extracted defect candidate parts by a first defect determination; and an inspection step of detecting harmful or harmless surface defects by a second defect determination using a convolutional neural network, the second defect determination being targeted at defect candidate parts after the screening by the first defect determination.
However, Totsuka does. (See figs. 7-8) In the same field of endeavors, the reference of Totsuka discloses and fairly suggests a screening step of screening the extracted defect candidate parts by a first defect determination (See fig. 8: S24; ¶ 65-76 where it teaches a defect classification unit that calculates at least one physical quantity for each detected defect candidate and compares the calculated physical quantity with a threshold value of the physical quantity to determine whether the defect candidate is a defect. This is a process of outputting defect candidates other than the defect candidate determined to be not to the defect classifier; only defect candidates other than the defect candidates determined not to be defects are the targets of defect classification using the neural network.); and an inspection step of detecting harmful or harmless surface defects by a second defect determination using a convolutional neural network, the second defect determination being targeted at defect candidate parts after the screening by the first defect determination. (See fig. 8: S25; ¶ 77-78 where it teaches a defect classifier configured to perform defect classification processing.)
Therefore, taking the combined teachings of Bian & Totsuka as a whole, it would have been obvious to one of ordinary skills in the art to incorporate these features into the system of Bian, in the manner as claimed and as taught by Totsuka, for the benefit of improving the accuracy of the defect determination. (See ¶ 51)
Re claim 2, the combination of Bian & Totsuka discloses wherein, in the first defect determination, feature values are extracted from the image, and the feature values are used for the screening. (In Totsuka, see ¶ 65-76)
Re claim 3, the combination of Bian & Totsuka discloses wherein in the first defect determination, a learning model is used for the screening, and the learning model is a model generated in advance by machine learning using feature values extracted from an image taken in advance of a surface of any inspection target. (In Bian, see fig. 4-5; ¶ 52-59, 66-67)
Re claim 4, Bian discloses a classification method of classifying surface defects on an inspection target (See fig. 4-9 where it teaches a neural network configured to detect defects in work piece.), the classification method including: an imaging step of acquiring an image of a surface of the inspection target (See fig. 1: 108; fig. 4: 302, 304; fig. 9: 902, 904; ¶ 20, 28, 43, 52; 86 where it teaches imaging device configured to obtain images of a work piece.); an extraction step of extracting defect candidate parts from the image. (See fig. 4; fig. 9: 906, 908; ¶ 55, 58-59, 63 where it teaches generating images 406 which represent defect candidates.)
But the reference of Bian fails to teach a screening step of screening the extracted defect candidate parts by a first defect determination; and a classification step of classifying types and/or grades of surface defects by a second defect determination using a convolutional neural network, the second defect determination being targeted at defect candidate parts after the screening by the first defect determination.
However, Totsuka does. (See figs. 7-8) In the same field of endeavors, the reference of Totsuka discloses and fairly suggests a screening step of screening the extracted defect candidate parts by a first defect determination (See fig. 8: S24; ¶ 65-76 where it teaches a defect classification unit that calculates at least one physical quantity for each detected defect candidate and compares the calculated physical quantity with a threshold value of the physical quantity to determine whether the defect candidate is a defect. This is a process of outputting defect candidates other than the defect candidate determined to be not to the defect classifier; only defect candidates other than the defect candidates determined not to be defects are the targets of defect classification using the neural network.); and a classification step of classifying types and/or grades of surface defects by a second defect determination using a convolutional neural network, the second defect determination being targeted at defect candidate parts after the screening by the first defect determination. (See fig. 8: S25; ¶ 77-78 where it teaches a defect classifier configured to perform defect classification processing.)
Therefore, taking the combined teachings of Bian & Totsuka as a whole, it would have been obvious to one of ordinary skills in the art to incorporate these features into the system of Bian, in the manner as claimed and as taught by Totsuka, for the benefit of improving the accuracy of the defect determination. (See ¶ 51)
Re claim 5, the combination of Bian & Totsuka discloses a management method, comprising a management step of classifying the inspection target, based on the types and/or the grades into which the surface defects have been classified by the classification method according to claim 4. (In Bian, see fig. 1, 4-9)
Re claim 6, the combination of Bian & Totsuka discloses a steel material production method, comprising: a production step of producing a steel material; and the inspection step included in the inspection method according to claim 1, wherein in the inspection step, the steel material produced in the production step is the inspection target. (In Bian, see fig. 1, 4-9; ¶ 26)
Re claim 7, the combination of Bian & Totsuka discloses a steel material production method, comprising: a production step of producing a steel material; and the management step included in the management method according to claim 5, wherein in the management step, the steel material produced in the production step is the inspection target. (In Bian, see fig. 1, 4-9; ¶ 26)
Re claim 8, the combination of Bian & Totsuka discloses a learning model generation method of generating a learning model to be used in an inspection method of detecting surface defects on an inspection target, the learning model generation method comprising a step of generating the learning model by a convolutional neural network using teaching data, which includes defect candidate parts after screening by a first defect determination on an image of an inspection target that has been acquired in advance as input record data, and results indicating whether the input record data is harmful or harmless as output result data. (In Bian, see fig. 4-5, 8; ¶ 60, 80 where it teaches training the neural network to detect defects in the work piece.)
Claim 9 has been analyzed and rejected w/r to claim 8 above.
Claim 10 has been analyzed and rejected w/r to claim 1 above.
Claim 11 has been analyzed and rejected w/r to claim 6 above.
Contact
Any inquiry concerning this communication or earlier communications from the examiner should be directed to LEON FLORES whose telephone number is (571)270-1201. The examiner can normally be reached M-F 8am - 6pm.
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, HENOK SHIFERAW can be reached at 571-272-4637. 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.
/LEON FLORES/Primary Examiner, Art Unit 2676 April 7, 2026
Prosecution Insights