US 11,861,871 B2
Crack evaluation of roofing membrane by artificial neural networks
Carl Wilfried, Wädenswil (CH); Katrin Schumann, Kriens (CH); Roger Kathriner, Kriens (CH); and Karin Odermatt, Sarnen (CH)
Assigned to SIKA TECHNOLOGY AG, Baar (CH)
Appl. No. 16/980,647
Filed by SIKA TECHNOLOGY AG, Baar (CH)
PCT Filed Mar. 28, 2019, PCT No. PCT/EP2019/057831
§ 371(c)(1), (2) Date Sep. 14, 2020,
PCT Pub. No. WO2019/185774, PCT Pub. Date Oct. 3, 2019.
Claims priority of application No. 18164450 (EP), filed on Mar. 28, 2018.
Prior Publication US 2021/0012159 A1, Jan. 14, 2021
Int. Cl. G06K 9/62 (2022.01); G06V 10/764 (2022.01); G06N 3/08 (2023.01); G06F 18/214 (2023.01); G06F 18/2415 (2023.01); G06V 10/82 (2022.01)
CPC G06V 10/764 (2022.01) [G06F 18/214 (2023.01); G06F 18/2415 (2023.01); G06N 3/08 (2013.01); G06V 10/82 (2022.01)] 16 Claims
OG exemplary drawing
 
1. A method for evaluating the crack intensity on a polymeric sheet based on a predetermined scale of crack intensity grades, comprising the steps of
a) recording a digital image of at least a portion of a surface of the polymeric sheet using an apparatus for recording digital images; and
b) automatic classification of the crack intensity by a computer-implemented program for pattern recognition by means of a trained artificial neural network, comprising
1) inputting the digital image or one or more subareas of the digital image to the trained artificial neural network as input data,
2) classification by the artificial neural network by assigning a grade from the predetermined scale of crack intensity grades to the digital image or the one or more subareas and
3) outputting the assigned grade or grades for the digital image and/or the one or more subareas as output data,
wherein the artificial neural network is trained in advance in a learning phase with a plurality of digital images or subareas thereof of polymeric sheet surface portions, whose grades in the predetermined scale are known and cover all grades of the predetermined scale and wherein the polymeric sheet is a roofing membrane or a sealing membrane,
wherein the learning phase comprises:
a training phase where a plurality of digital images or subareas thereof are input to the artificial neural network and the artificial neural network is provided with the known grade of each of these digital images as feedback, and
a test phase where the artificial neural networks classifies a plurality of digital images or subareas thereof and the grades assigned by the artificial neural networks are compared with the known grades to determine a matching probability, and
optionally repeating the training phase and the test phase until the matching probability desired is reached.