Systems and methods for quantifying concrete surface roughness

Patent metadata

Field	Value
Publication number	US-11468552-B1
Application number	US-202117373057-A
Country	US
Kind code	B1
Filing date	Jul 12, 2021
Priority date	Jul 12, 2021
Publication date	Oct 11, 2022
Grant date	Oct 11, 2022

Abstract

The degree of concrete surface roughness contributes to the bond strength between two concrete surfaces for either new construction or repair and retrofitting of concrete structures. Provided are novel systems and methods with industrial application to quantify concrete surface roughness from images which may be obtained from basic cameras or smartphones. A digital image processing system and method with a new index for concrete surface roughness based on the aggregate area-to-total surface area is provided. A machine learning method applying a combination of advanced techniques, including data augmentation and transfer learning, is utilized to categorize images based on the classification given during the learning process. Both methods compared favorably to a well-established method of 3D laser scanning.

First claim

What is claimed is:

1. A system for determining a measurement of surface roughness of a concrete sample, the system comprising:
a processor; and
a machine-readable medium in operable communication with the processor and having instructions stored thereon that, when executed by the processor, perform the following steps:
receiving training images;
using a data augmentation technique to train a convolutional neural network by performing one or more random left rotation of the training images, random right rotation of the training images, changing a brightness of the training images, blurring the training images, horizontal flipping of the training images, vertical flipping of the training images, and resizing of the training images;
measuring and labeling surface roughness values of the training images using at least one of the 3D laser scanning and digital image analysis;
using the label surface roughness values to train the convolutional neural network;
receiving an image of the concrete sample;
defining a positive integer n and a positive integer index i ranging from 1 to n;
defining a set of n roughness classes (Ci);
defining for each Ci an associated average roughness value (Ci av );
generating for each Ci, a probability (P i ) of matching the image with that respective Ci, the generating of the P i comprising using the trained convolutional neural network; and

determining a weighted average roughness value (R a ) for the image from the sum of each P i multiplied by the respective Ci av to obtain the measurement of surface roughness of the concrete sample.

2. The system according to claim 1 , the determining of the R a for the image comprising using the following equation:

R
a

=

∑

i
=
1

n

(

(

P
i

)

·

(

Ci

a
⁢
v

)

)

.

3. The system according to claim 1 , the data augmentation technique being applied in an offline manner to increase the sample size of training data.

4. The system according to claim 1 , the convolutional neural network being trained using a transfer learning technique.

5. A system for determining a measurement of surface roughness of a concrete sample, the system comprising:
a processor; and
a machine-readable medium in operable communication with the processor and having instructions stored thereon that, when executed by the processor, perform the following steps:
receiving training images;
using a data augmentation technique to train a convolutional neural network by performing one or more random left rotation of the training images, random right rotation of the training images, changing a brightness of the training images, blurring the training images, horizontal flipping of the training images, vertical flipping of the training images, and resizing of the training images;
measuring and labeling surface roughness values of the training images using at least one of the 3D laser scanning and digital image analysis;
using the label surface roughness values to train the convolutional neural network;
receiving an image of the concrete sample;
defining a positive integer n and a positive integer index i ranging from 1 to n;
defining a set of n roughness classes (Ci),
defining for each Ci an associated average roughness value (Ci av );
generating for each Ci, a probability (P i ) of matching the image with that respective Ci, the generating of the P i comprising using the trained convolutional neural network; and and
determining a weighted average roughness value (R a ) for the image from the sum of each P i multiplied by the respective Ci av to obtain the measurement of surface roughness of the concrete sample;

the determining of the R a for the image comprising using the following equation:

R
a

=

∑

i
=
1

n

(

(

P
i

)

·

(

Ci

a
⁢
v

)

)

;

the convolutional neural network being further trained using a transfer learning technique; and
the data augmentation technique applied in an offline manner to increase the sample size of training data.

Assignees

• Valikhani Alireza
• Jahromi Azadeh Jaberi
• Pouyanfar Samira
• Mantawy Islam Mohamed
• Azizinamini Atorod
• The Florida International Univ Board Of Trustees

Inventors

• Valikhani Alireza
• Jahromi Azadeh Jaberi
• Pouyanfar Samira
• Mantawy Islam Mohamed
• Azizinamini Atorod

Classifications

• G06T2207/30132

  Masonry; Concrete · CPC title

• G06T7/0004 Primary

  Industrial image inspection · CPC title

• G06T2207/20084

  Artificial neural networks [ANN] · CPC title

• G06T2207/20081

  Training; Learning · CPC title

• G06F18/2415

  based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate · CPC title