Image Analysis of Concrete to Detect the Potential of a Sequential

Rift

Fahmi, Jhoni Hidayat and Suherman

Department of Electrical Engineering, Faculty of Engineering, University of Sumatera Utara,

Jl. AlmamaterKampus USU Medan 20155 INDONESIA

Keywords: Crack detection, sequential, grayscale, median filter, edge detection, threshold, subtracted image.

Abstract. Concrete cracks in the construction world is a common phenomenon that occurs in all types of concrete

structures. Due to our ignorance of damage to concrete structures, small cracks are often ignored. In fact, a

small crack can cause a big disaster. Therefore, early detection of cracks in concrete itself is very important

that is expected to minimize the disaster caused. In detecting cracks of the concretesurface,a structure is

done by several methods. The purpose of this research is to detect the potential of cracking on the surface of

the concrete image with sequential digital image processing. In this research, crack detection is applied to

the concrete surface of concrete test result done by applying grayscale function, noise filter, edge detection,

threshold, followed by comparing the sequential image with image subtract method. Based on the research

data, the methods applied are able to detect the potential of crack in the sequential image with the

percentage of the result of True Positive Rate equal to 76.2% and False Negative Rate of 23.8%. The

applied methods can be used as one of the approaches to detect the potential of cracking on the surface of

concrete images in concrete press concrete sequentially.

1 INTRODUCTION

In the construction world, concrete is a composite

building material made from a combination of

aggregates and cement binders. In a concrete

structure, cracking is one of the serious problems

(Nishikawa, 2012). Although cracks are on the

surface, it is difficult to be visually detected if the

crack is small (Broberg, 2013). Each crack model

will result in damage and failure of various models

on concrete structures (Litorowitz, 2006). Each

concrete structure will surely have a crack, which

needs to be noticed is whether the crack can be

tolerated and is not at risk or cracks that harm the

overall structure of the building or not. Some of the

techniques used to check for cracks in concrete

include electron microscope scanners, fluorescent

optical microscopes. Percolation technique based on

liquid-based infiltration phenomena is based on

image processing (Yamaguchi, 2010). In recent

years, Micro-Computed X-ray Tomography (Micro-

CT) has emerged as a characterization tool because

of its ability to provide non-destructive three-

dimensional high-resolution images (Yu, 2016).

In this study, in detecting the potential of

sequential rupture, the authors applied a method

based on previous research on digital image

processing to detect cracks, then using subtract

image method to detect the potential of the crack.

2 MATERIAL AND METHOD

The implementation of the research begins by

identifying the problem and establishing the research

objectives. In the Research Design the Median Filter

method, Sobel edge, Otsu Threshold, Filter Area,

and Image Subtract Method are used. The observed

crack is a crack image that occurs on the specimen

when loaded by a compressive test machine. The

observed variables are image changes that occur

from the image comparison results during the test.

Sample data collection is done based on the

problem solved. The calculation of the sample data

test is required to obtain the performance results of

the applied method and draw conclusions.

Fahmi, ., Hidayat, J. and Suherman, .

Image Analysis of Concrete to Detect the Potential of a Sequential Rift.

DOI: 10.5220/0008881900310035

In Proceedings of the 7th International Conference on Multidisciplinary Research (ICMR 2018) - , pages 31-35

ISBN: 978-989-758-437-4

2.1 Data Collection Process

The sample of concrete surface observed was

cylindrical concrete with diameter 15 cm, height 30

cm with concrete quality FC 41,5 (500 Kg / cm2).

FC 37.35 (450 Kg / cm2), FC 33.2 (400 Kg / cm2),

FC 29.05 (350 Kg / cm2) which is the material

available at the time of the study.

The data taken is the concrete surface image of

the compressive test, using Canon 60D DSLR

Camera with auto mode, equipped with external

tripod and shutter by sampling 1 frame in 2 seconds.

Data collection is done in the morning until the

afternoon from 10:00 to 15:00 pm in a laboratory

room with sunny weather conditions. The data taken

is expected to be static for each frame. The distance

between the camera and the test object is about 1

meter with sufficient lighting levels. The data

collection process is shown in Figure 1.

(a) (b)

(c)

Figure 1: Data collection process (a) Thespecimen put

into the test machine, (b) The process of loading the load

onto the specimen, (c) The process of retrieving the image

of the specimen.

2.2 Data Processing

Conducted data entry in the form of image t

and

image t

arethen processed through Processing Citra

with the application of functions and methods used

for data processing. Performingcomparison of pixels

per pixel between image t

with image t, to find any

change in both images with a predetermined

threshold is done. If there is no change then re-

infiltrated is conducted until the final image, or until

the image changes above the specified threshold.

Then the changed image is saved and displayed. The

Data Processing Process can be illustrated through

the workflow diagram in Figure 2.

Figure 2: Diagram of the stage of data processing.

Figure 3: Diagram of image processing.

ICMR 2018 - International Conference on Multidisciplinary Research

2.3 Image Processing

From Sattar's research (Sattar,2016), in his

Processing method it is seen that the input image in

cropping or image cropping process at certain

coordinate is considered as a workarea and then

converted into a grayscale image, then re-processed

by noise filter that is by Median Filter method to

reduce noise on the image; then processed again for

edge detection using Sobel Filter, continued to Otsu

Thresholding with Filter Area. Image Processing

Process Diagram is shown in Figure 3.

2.4 Input Image

The input image is a digital image of a concrete

surface photo with a color image kernel (RGB) in

JPEG format with dimensions of 5184 x 3456

pixels. The input image is shown in Figure 4.

Figure 4: Input Image.

2.5 Grayscale Image

A grayscale image is a process done on a digital

image to change the original color of the image into

gray color. Input image in the form of the digital

image has 3 colors in each pixel: Red, Green, and

Blue, each with a different value. An illustration of

the RGB pixel values in the input image is shown in

Figure 5 and Figure 6.

Figure 5: The RGB value of the input image.

Figure 6: Grayscale changes in the input image.

2.6 Otsu Threshold

The next step is to determine the threshold using this

Otsu method. The Otsu method divides the image

into two groups, namely: target pixels and

background pixels (Sattar,2016 and Huang, 2015).

This method selects a threshold value based on the

minimization of intra-class variants (the variants in

the class). Minimizing intra-class variants is the

same as maximizing inter-class variants (Otsu 1979).

2.7 Measurement Parameters

The measurement of the performance results of the

method uses two parameters. The first parameter is

the True Positive Rate, i.e., the ratio of the number

of data features successfully detected, exactly

divided by the total number of sample data tested.

The second parameter is the False Negative Rate,

which is the ratio of the amount of data that is wrong

or failed to be detected by the total number of

sample data tested. Performance measurement

results of this method will be the conclusion of how

the success rate of the system is applied in research.

3 RESULTS AND DISCUSSION

The data used as the test sample are the drawings of

21 cylindrical concrete with the quality of FC 41.5,

FC 37.35, FC 33.2, FC 29.05 obtained from

laboratory tests taken sequentially with digital

cameras.Each concrete data has 15 or more images

taken in sequential. Before the image comparison

process is done, the sample data is a cropping

process or image cutting process at a certain

coordinate which is considered as a work area; then

proceeded to the image processing consisting of

grayscale process on the original image, filtering on

the grayscale image by using a median filter, edge

detection using Sobel and Otsuthreshold and filter

area to sample data shown in figure 7.

Image Analysis of Concrete to Detect the Potential of a Sequential Rift

The test is performed on 21 test data, wherein

each test data has a sequential image data image

where objects with the number of an area less than

700 pixels are detected, considered as noise, then the

objects are removed. The value of the first test result

is obtained, considered an image that has the

potential to crack.

Table 1 and Table 2 show the test results of some

concrete by comparing the image after the image

before sequentially using the subtract image method

and the addition of the area filter function.

Figure 7: Results Process Image Processing(a) the

original image, (b) cropping the image for the work area,

(e) the filter result image, (f) the Otsu threshold image, (g)

filter area results.

Table 1: Results of concrete testing 1.

Concrete Time

Number of

Pixels

Image

Subtract

Testing with

area filter>

700

Con

crete 1

t0 887 - 0

t1 1103 1631 0

t2 971 1395 0

t3 1816 2751 0

t4 983 867 0

t5 1115 1576 0

t6 1116 1607 0

t7 1106 1470 0

t8 898 1127 0

t9 1020 1413 0

t10 754 1000 0

t11 875 1209 0

t12 1672 2867 1299

t13 1436 1053 0

t14 1270 1190 0

Table 2: Results of concrete testing 4.

Concrete Time

Number of

Pixels

Image

Subtract

Testing

with area

filter> 700

Con

crete 4

t0 2152 - 0

t1 3154 5258 0

t2 3400 5352 0

t3 2964 4413 0

t4 2973 5159 0

t5 2913 4730 0

t6 2447 4081 0

t7 2897 5230 0

t8 31421 64865 45068

t9 40358 64307 18133

t10 32173 44134 2522

t11 33265 50945 8938

t12 40813 59743 7734

t13 38260 51123 2385

t14 38299 54510 6793

t15 27825 39672 0

Table 3: Concrete Test Result.

Concrete

Total

Image

Test Result Description

1 15 True

2 15 True

3 15 False

The system does not

detect any cracks, but

cracks are seen in the

image t=13

4 16 True

5 16 True

6 16 True

7 16 False

The system detects a

crack in the image

t=5, but no visible

cracks

8 16 True

9 18 True

10 17 False

11 18 True

12 18 False

The system detects a

crack in the image t =

8, but the crack is seen

in the image t = 6

13 15 False

The system does not

detect any cracks, but

the crack is seen in the

image t = 12

14 15 True

15 18 True

16 18 True

17 18 True

18 15 True

19 18 True

20 18 True

21 18 True

ICMR 2018 - International Conference on Multidisciplinary Research

Based on Table 3 of the test results on 21

concrete, the results show the success of the

succession shown in Table 4

Table 4: Percent of Success.

Amount of

sample data

Number of

Testing True

True

Positive

False

Negative

21 16 76,2 % 23,8 %

From Table 4 based on the test data taken, the

result of the applied method is the ability to detect

precisely the potential of cracking in sequential

image with the percentage of result of True Positive

Rate equal to 76,2% and False Negative Rate equal

to 23,8%, where from 21 concrete tested, the number

of tests considered correct is 16 concrete.

4 CONCLUSION

From the research results obtained, the

conclusionsare as follows:

1. Methods in the research can be used as a

potential approach to the detection of cracks in

the image data sequentially concrete

compression test results.

2. From the research data, the applied method

resulted in a True Positive Rate is 76.2% and

from the 21 concrete tested, the correct test

amount is 16 concrete.

3. The method used can be applied to research

data with good image quality, adequate lighting

and static image of each frame of data retrieval

during research.

4. The success rate of detection depends on the

image capture process, as well as the quality of

the test image. The image quality and

illumination of the bad image will influence the

research result.

REFERENCES

B Yu, R S Bradley, C Soutis, and P J Withers, 2016. A

comparison of different approaches for imaging cracks

in composites by X-ray microtomography, in Royal

Society Publishing,

Broberg P, 2013. Surface crack detection in welds using

thermography, in NDT&E International, vol. 57,

pp.69-73.

Dorafshan Sattar , Maguire Marc , and Qi Xiaojun, 2016,

Automatic Surface Crack Detection in Concrete

Structures Using OTSU Thresholding and

Morphological Operations, in CEE Faculty

Publications.

EkaDodiSuryanto, 2015,

EkstraksiFiturHaralickMenggunakan Citra Mikroskop

Digital TrinocularUntuk Proses

IdentifikasiCacingPenyakit Kaki Gajah, in

TeknikElektro, FakultasTeknik, Universitas Sumatera

Utara.

Litorowicz A, 2006, Identification and quantification of

cracks in concrete by optical fluorescent microscopy,

in Cement and Concrete Research, vol. 36, pp. 1508-

1515

Nishikawa T, J Yoshida, T Sugiyama, and Y Fujino, 2012.

Concrete Crack Detection by Multiple Sequential

Image Filtering, in Computer-Aided Civil and

Infrastructure Engineering, vol. 27, pp. 29-47.

Sony NuryadinSyarifuddin, 2006, Analisis Filtering Citra

DenganMetode Mean

Filter dan Median Filter, in JurusanTeknikInformatika,

FakultasTeknikdan

IlmuKomputer, UniversitasKomputer Indonesia,.

Yamaguchi Tomoyuki and Shuji Hashimoto, 2010, Fast

crack detection method for large-size concrete surface

images using percolation-based image processing, in

Machine Vision and Applications, pp. 797 809.

Zhangcan Huang, Fan Xi, Liu HaiMing A.M.A. Talab,

2015, Detection crack in the image using Otsu method

and multiple filtering in image processing techniques,

in Optik - Int. J. Light Electron Opt.

Image Analysis of Concrete to Detect the Potential of a Sequential Rift