Image Distance Measurement with only Camera using OpenCV

Object Detection

Friendly

, Harizahayu

and Rahmat Widya Sembiring

Computer Engineering, Politeknik Negeri Medan, Sumatera Utara, Indonesia

Informatica Management, Politeknik Negeri Medan, Sumatera Utara, Indonesia

Keywords: Image Processing, OpenCV, Machine Learning, Distance Measurement.

Abstract: Image processing combine with machine learning is used widely in image recognition, image classification,

and image detection. By using camera to detect an object has been done within several article by Guamán &

Naranjo and Shah S.P.. The other usage of image and object detection is in measurement of distance. Distance

measurement is possible by using additional peripheral such as sensor and extra camera. In previous research

by S.H.Chen, detecting an object by knowing focal length of camera and object using spatial dan disperse

technic can produce error percentage of max 5.72%, and deviation of 0.79. The distance used is in meter.

Thus the deviation is in 79cm. This research is intended to elaborate the usage of single camera in detecting

and measuring distance without using any sensor and can be used without knowing the focal length of a

camera. The distance to be measured is between the camera and the object. In this research, author used a

square object with different size and different distance range. By using one of the known data as a pivot in

calculating other image distance, the average error between 4%-7% . These result was achieved by using

different object size. The bigger the size of the object used as reference, the smaller the error percentage of

the measurement.

1 INTRODUCTION

Measurement is the act of determining a target's size,

length, weight, capacity, or other aspect. The distance

between an object is determined by using

measurement. Distance measurement can be done by

using ruler, sensor, and any other devices. In this

journal, author try to find a way to measure distance

by using only camera.

The usage of camera is more common in daily

live as camera is already part of gadget around us.

Camera can be found in almost everyday devices like

smartphones and laptop. Many research has been

done in order to determine the distance between

object and camera (Guamán & Naranjo, 2019;

Dirgantara, Rohman, & Yulianti, 2019) (Valocký,

Drahoš, & Haffner, 2020). One of the method is by

using object classification and machine learning. The

most common library for image classification is

OpenCv (Gupta, 2017).

In order to help people in extending the function

of a camera, this research try to formulate a

measurement and experiment the parameter and

formula to calculate the distance between a camera

and an object. By using only the camera dan method

propose, the camera can be used for measurement.

This research will used the image classification

and edge detection in detecting the distance between

camera and object. The library that is used are

OpenCv. This research is used in measuring short

distance between 20cm to 200cm.

2 RELATED WORKS

One of the research state how the modified camera

and with the help of sensor can enhanced distance

measurement (Valocký, Drahoš, & Haffner, 2020). In

research by Valocký, Drahoš, & Haffner, the distance

measurement is used for object measurement

detection in range between 110cm and 163cm with

average error of 41µm. The research is about

measuring distance between object and pattern.

While Shi-Huang Chen in his research (Chen &

Chen, 2011) used distance measurement using

camera with car license plate as object reference. In

Shi-Huang Chen research, the method used is the

method of triangular and founding the focal length of

774

Friendly, ., Harizahayu, . and Widya Sembiring, R.

Image Distance Measurement with only Camera using OpenCV Object Detection.

DOI: 10.5220/0010953300003260

In Proceedings of the 4th International Conference on Applied Science and Technology on Engineering Science (iCAST-ES 2021), pages 774-778

ISBN: 978-989-758-615-6; ISSN: 2975-8246

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

camera. The distance measured are between 13m-

31m. The image recognition used are with image

capture in range of 5, 8, 11, 14, 17, 20, 23, 26, 29, 32,

35, 38, 41, 44, 47, and 5. The research data present in

Shi-Huang Cen article is present in the following

table.

Table 1: Recapitulation Of Distance Result Measurment

(Chen & Chen, 2011).

Real

Distance

Sparse Method Dense Method

Calculated

Distance

Erro

Devia

tion

Calculated

Distance

Erro

Devia

tion

13,8 14.59 5,72 0.79 13,81 0,07 0,01

20,6 21,39 3,83 0.79 20.,97 1,8 0,37

22,7 23,28 2,56 0.58 23 1,32 0,3

23,2 23,7 2,16 0,5 23,46 1,12 0,26

29,9 30,29 1,3 0,39 30,68 2,61 0,78

32,2 31,77 1,34 0,43 32,32 0,37 0,12

From the data presented, the measurement is set

in meter, thus the deviation are around 39 cm to 79

cm. The shortest the distance, the bigger the

deviation.

In other research to measure distance like

(Dirgantara, Rohman, & Yulianti, 2019) and (Gao,

Chen, Liu2, & Chen2, 2021) used the same method

by using references. In both paper state that by using

bottom part of a vehicle image, measuring the

distance between the two object can be achieved

though it state that it’s not accurate in a certain length.

As previous test result shown which is not clear, but

the average of error percentage are 1% and deviation

of 26 cm. Both research using mobile net and Yolo in

detecting image. While Jae Moon Lee (Lee, Hwang,

& Jung, 2021) in their research, the object detected by

using difference between 2 image with same object,

where there is a movement distance between 2 image.

2 RESEARCH DESIGN

Implementation of image detection is common

especially in image processing.

Figure 1: Image Capturing process in camera.

The process of measuring in this research is

following the common concept of image capturing in

camera lens. The image illustration is describe in

Figure 1.

This research will used 2 different object to

calculate the distance of the object. The first object is

16cm x 16 cm and 20cm x 20cm. The method propose

in this paper are to create a comparison between the

measured distance to get the ideal pivot used for

measurement.

Different from the process of captured image by

the camera, this research will approach by using the

concept of triangle. This deduction come from the

perspective of image we get from the capture image.

The further the distance of an object to camera, the

smaller the image captured. Figure 2 can describe the

process of image captured.

Figure 2: Image Capturing illustration.

The image shown how The further the distance of

an object to camera, the smaller the image result.

Despite the angel of the camera, the image proportion

will remain as long as the direction where the image

taken is equal.

To produce the formula, we must assume that the

angle at α in Figure 1 is equal both at real distance and

image distance. If we assume that the α is a formula

of tangent, then the formula can be described as:

tan(



) =



∶𝐷





∶ 𝐷



(1)

Thus

𝐷



𝐷



𝑥ℎ

𝐻

(2)

By using this formula, we can have a base

distance. As we follow the rules of triangle similarity,

as the image move further, so does the distance and

the image grow smaller. We can use one of the image

captured and with known distance as a reference. By

using this as a reference, we can derive the formula

by comparing the distance inside the reference image

) and the height of the measured image captured.

The formula for calculated distance D

is calculated

using formula as follow:

Image Captured

Image Distance Measurement with only Camera using OpenCV Object Detection

775

𝐷



𝐻 𝑥 𝐷



ℎ



(3)

To calculate the distance inside the image, this

research will be using this formula.

2.1 Algorithm

The algorithm used in testing the method in this

research will be using the opencv image classification

and then use an image as reference and calculated the

error percentage.

The process is started by collecting and arranging

the imaged into array. The image thus choose one by

one to be used as a reference. For each reference, the

formula will be calculated against other data. The

calculated distance is then compared to the image real

distance. The errors percentage and the deviation thus

calculated to produce the report.

The algorithm proposed is as describe in figure 3.

Figure 3: Algorithm Of Distance Measurement Testing.

3 EXPERIMENTAL RESULT

In this paper, the experiment is created by using

python programming language. The library used is

OpenCv. In experimenting the method, the object

used is determined and experiment used 2 object as

reference. The first one is a square paper with width

of 16cm x 16 cm and square paper with the width of

20cm x 20cm. The object’s distance thus measured

with ruler and the picture was taken by camera. The

distance is arranged in multiplication of 10. The

number of sample taken are 20 for each object. Some

image of the object taken shown in figure 3.

(a)

Image 1 taken from

10 cm distance

(b)

Image 1 taken from

100 cm distance

(c)

Image 2 taken from

10 cm distance

(d)

Image 2 taken from

100 cm distance

Figure 4: Image taken from different distance.

The process of calculating distance is done by

iterating the image inside the dataset. The process

include the image masking, image detection and

measuring the area of the reference object. Since the

object used is a square, the side of the object can be

calculated as a square root of the area. Thus the side

of the object can be calculated easily. After the

calculation of the sides of the object, the calculation

will continue to calculated the distance by comparing

the imaginary Distance and then reference distance

and will produce the real distance. Since the

information of the dataset is available, the process

continues to calculate the deviation and the error of

the measurement. The calculated data will be written

in the image and shown. The simulation is shown in

Figure 5.

Calculate the error percentage and

deviation

start

Collect and arrange

ima

e data in I

and I

Choose next image

as reference in I

Calculate the D

from

reference image

Calculate the Real Distance base

from reference image

Get next ima

Next image

available?

Finish

No more ima

Yes

No more

image

Yes

Next image

ava

iCAST-ES 2021 - International Conference on Applied Science and Technology on Engineering Science

776

Figure 5: The Process Of Simulation.

The test result of the simulation can be seen in

table 2 and table 3. Table 2 shown the testing result

of simulation by using object reference of square with

dimension of 16 cm x 16 cm. Table 3 shown the

testing result of simulation by using object reference

of square with dimension of 20 cm x 20 cm.

Table 2: Test Result Using Object Reference of Square

Object of 16 cmx 16 cm.

Error

(

)

Deviation

(

in cm

)

Avg Max Min Avg Max Min

10 76.59 99.03 38.6 48.48 99.51 5.57

20 30.19 43.6 9.59 27.04 60.72 2.63

30 20.54 34.17 7.08 19.85 47.36 1.92

40 14.56 38.52 6.18 14.25 36.55 2.12

50 10.22 42.1 0.22 9.26 26.45 0.13

60 10.08 42.23 0.22 9.08 26.06 0.11

70 8.91 43.58 2.34 7.28 21.9 1.31

80 8 44.9 1.01 5.61 17.62 0.9

90 7.72 45.45 0.69 4.97 15.79 0.69

100 7.6 45.83 0.69 4.59 14.51 0.62

110 7.53 46.78 0.1 3.86 11.18 0.13

120 7.51 46.47 0.48 4.04 12.27 0.63

130 7.52 46.73 0.1 3.88 11.36 0.11

140 7.7 47.19 0.4 3.78 9.72 0.6

150 7.84 47.4 0.37 3.79 9.01 0.56

160 7.99 47.6 0.04 3.86 9.08 0.08

170 8.17 47.75 0.26 4.01 9.14 0.46

180 8.01 47.62 0.04 3.88 9.09 0.07

190 8.42 47.93 0.35 4.26 9.21 0.59

200 11.28 49.76 3.5 7.5 9.9 6.51

In order to eliminate unnecessary decrease int

error percentage and deviation, the test result using

reference image against the reference image itself is

remove since the test result always shown no

deviation and no error.

Table 3: Test Result Using Object Reference of Square

Object of 20 cmx 20 cm.

Error (%) Deviation (in cm)

Max Min Av

Max Min

20 27.66 39.52 12.07 25.49 54.19 3.23

30 14.99 24.5 8.59 15.77 36.6 2.41

40 7.48 17.82 2.19 8.29 22.57 1.07

50 5.82 19.58 1.53 6.3 18.69 0.87

60 4.05 22.21 0.08 3.69 13.37 0.06

70 4.06 22.15 0.05 3.73 13.5 0.05

70 4.04 22.33 0.02 3.63 13.12 0.02

80 4.27 21.65 0.33 4.12 14.48 0.33

90 4.83 20.8 0.76 4.96 16.16 0.75

100 4.42 21.39 0.33 4.35 14.99 0.26

110 4.04 22.35 0.02 3.62 13.09 0.01

120 4.07 22.1 0.05 3.75 13.58 0.04

130 4.1 22.63 0.36 3.55 12.52 0.26

140 4.28 23.16 0.07 3.5 11.44 0.11

150 4.3 23.21 0.07 3.5 11.33 0.1

160 4.36 23.3 0.12 3.53 11.14 0.18

170 8.5 28.33 1.61 7.12 11.22 2.94

180 7.17 27.16 0.06 5.54 8.46 0.11

190 7.12 27.11 0.02 5.49 8.36 0.04

200 7.11 27.1 0.02 5.48 8.33 0.04

From the result table can be seen that the deviation of

distance measurement is highest when the reference

is using smaller image reference. As the pixel size

tend to grow small, the accuracy is increased. In the

Table 2, the accuracy is the highest when using the

object reference of 120cm while using the image

larger in Table 3 shown that the accuracy getting

better result when the object reference is 110cm. The

deviation of measurement is decreasing as the object

reference is higher. This happen in both table.

Image Distance Measurement with only Camera using OpenCV Object Detection

777

4 CONCLUSIONS

By using comparison method and triangular

similarity, this research shown a great accuracy of

96% or 4.04% error. The deviation of measurement

achieved from this experiment is average of 3.02%.

Although some of the measurement result shown

small deviation but against other image captured

shown greater deviation. The irregular result in this

research show that the method and formula need to be

improve. But comparing to greater deviation in other

research, this method can be shown in less needed

accuracy measurement. By using the calculation

formula of:

𝐷



𝐻 𝑥 𝐷



ℎ



(3)

We can calculate the distance between object only

using object references. Comparing this result with

other method in other research cited in this research,

which computing and detecting complex image,

measuring the distance by using determined object

can result to much better accuracy in short distance

and greater for further distance. The downside of this

method is that the object reference used must be

varied according to the distance as smaller object can

not be detected in further distance.

ACKNOWLEDGEMENTS

This work was supported and funded by Politeknik

Negeri Medan PDK Program for publication and

research.

REFERENCES

Chen, S.-H., & Chen, R.-S. (2011). Vision-Based Distance

Estimation for Multiple Vehicles Using Single Optical

Camera Feature. Second International Conference on

Innovations in Bio-inspired Computing and

Applications. Kaohsiung.

Cyganek, B. (2013). Object Detection And Recognition In

Digital Images. Chichester: John Wiley & Sons, Ltd.

Dirgantara, F. M., Rohman, A. S., & Yulianti, L. (2019).

Object Distance Measurement System Using

Monocular Camera on Vehicle. International

Conference on Electrical Engineering, Computer

Science and Informatics (EECSI 2019). Bandung.

Gao, W., Chen, Y., Liu2, Y., & Chen2, B. (2021). Distance

Measurement Method for Obstacles in front of Vehicles

Based on Monocular Vision. 2020 2nd International

Conference on Computer, Communications and

Mechatronics Engineering (CCME2020). Xiamen,

China: IOP.

Guamán, L. R., & Naranjo, J. E. (2019). Object detection in

rural roads through Single Shot Multibox Detector

Mobilenet network. Proceedings of 2019 the 9th

International Workshop on Computer Science and

Engineering. Hong Kong.

Gupta, B. &. (2017). Study on Object Detection using Open

CV - Python. International Journal of Computer

Applications, 17-21.

Lee, J. M., Hwang, K., & Jung, I. H. (2021). Real Distance

Measurement Using Object Detection of Artificial

Intelligence. Turkish Journal of Computer and

Mathematics Education Vol.12 No.6(2021), pp. 557-

563.

Sasaki, N., Tomaru, S., & Nakamura, S. (n.d.).

Development of Inter-Vehicle Distance Measurement

System using Camera-Equipped Portable Device. 2017

17th International Conference on Control, Automation

and Systems. Jeju, South Korea.

Shah, S. P. (2021). A Motion Detection System in Python

and Opencv. International Conference on Intelligent

Communication Technologies and Virtual Mobile

Networks (ICICV), pp. 1378-1382.

Valocký, F., Drahoš, P., & Haffner, O. (2020). Measure

distance between Camera and Object using Camera

Sensor. Cybernetics & Informatics (K&I), 1-4.

iCAST-ES 2021 - International Conference on Applied Science and Technology on Engineering Science

778