UNIVERSITAS
KRISNADWIPAYANA
Jakarta - Indonesia,
22 December
2022
The 2022 2nd
International Seminar on Machine Learning, Optimization, & Data Science
ismode.unkris.ac.id
Stereo Camera Calibration For
Autonomous Car Applications
Muhammad Abdillah Rahmat, Indrabayu, Andani Achmad, Ejah
Umraeni Salam, dan Muhammad Fadhil Bin Bahrunnida
Introduction
Problem
 The development of Indonesian
transportation sector is very high.
 Intelligent Transportation System
(ITS) is a vastly developing field in
the world of transportation.
Most object detection research still
uses one camera
 How to implement calibration in-
camera stereo using a webcam
Proposed Solution
 Stereo camera calibration
for object detection
implementation
Literature Study
The research conducted by Abdelmoghit Zarane
et al. with the title Distance measurement
system for autonomous vehicles using stereo
camera.
1
The research conducted by Yasir Dawood Salman
et al. with the title of Distance Measurement For
Self-Driving Cars Using Stereo Camera.
2
System Stereo Design
The study used 2 Logitech webcams; c920 stream pro and c922 stream pro for camera calibration.
Logitech Webcam c922 stream pro
Logitech Webcam c920 stream pro
Stereo camera system stages
Stereo
Calibration
Stereo
Rectification
Stereo Match Triangulation
The steps taken to calibrate the c920 stream pro and c922 stream pro webcams in
this study are :
• Sets the number of angles in and the size of the chessboard pattern as well as the distance between the
lens and the shooting distance.
• Find the coordinates of each corner of the chessboard
• Find the coordinator of each corner of the chessboard at sub-pixel points
• Finding intrinsic and extrinsic values through each coordinate
Sets the number of angles in and the size of the chessboard pattern
as well as the distance between the lens and the shooting distance.
• The number of angles in the chessboard pattern
• The unit size of each pattern
• The distance between the camera lens to be stereotyped
• The shooting distance between the stereo camera and the chessboard pattern
Chessboard patterns used
1
Find the coordinates of each corner of the chessboard
The next step is to find the coordinates of each corner point on the chessboard using the cv2.findChessboardCorners
function.
In using this function 3 supporting arguments are needed :
• The source of the image
• The number of inner corners in the image
• The argument of the operation flags.
2
Find the coordinator of each corner of the chessboard at sub-pixel
points
In the calibration process, accuracy and precision are the most decisive values. To obtain an accurate result, it is
necessary to obtain the location of the angle with an accuracy level of up to sub-pixels.
3
Finding intrinsic and extrinsic values through each coordinate
The final step in the calibration process is to pass the 3D point on the real-world coordinates and the 2D location
of each coordinate in all images to the cv2.calibrateCamera function.
4
Result
RMSE value with chessboard with a pattern unit size of 3 cm
RMSE value with chessboard with a pattern unit size of 11
cm
In building a stereo vision system, the RMSE value is output by the cv2 function. Calibrate stereo becomes an
evaluative foundation in the calibration process. In this study, two types of chessboard patterns were used with
several shooting scenarios for camera calibration needs. Table 1 and Table 2 show the RMSE values obtained from
each shooting scenario for calibration needs
Thank You.

Stereo Camera Calibration For Autonomous Car

  • 1.
    UNIVERSITAS KRISNADWIPAYANA Jakarta - Indonesia, 22December 2022 The 2022 2nd International Seminar on Machine Learning, Optimization, & Data Science ismode.unkris.ac.id Stereo Camera Calibration For Autonomous Car Applications Muhammad Abdillah Rahmat, Indrabayu, Andani Achmad, Ejah Umraeni Salam, dan Muhammad Fadhil Bin Bahrunnida
  • 2.
    Introduction Problem  The developmentof Indonesian transportation sector is very high.  Intelligent Transportation System (ITS) is a vastly developing field in the world of transportation. Most object detection research still uses one camera  How to implement calibration in- camera stereo using a webcam Proposed Solution  Stereo camera calibration for object detection implementation
  • 3.
    Literature Study The researchconducted by Abdelmoghit Zarane et al. with the title Distance measurement system for autonomous vehicles using stereo camera. 1 The research conducted by Yasir Dawood Salman et al. with the title of Distance Measurement For Self-Driving Cars Using Stereo Camera. 2
  • 4.
    System Stereo Design Thestudy used 2 Logitech webcams; c920 stream pro and c922 stream pro for camera calibration. Logitech Webcam c922 stream pro Logitech Webcam c920 stream pro
  • 5.
    Stereo camera systemstages Stereo Calibration Stereo Rectification Stereo Match Triangulation
  • 6.
    The steps takento calibrate the c920 stream pro and c922 stream pro webcams in this study are : • Sets the number of angles in and the size of the chessboard pattern as well as the distance between the lens and the shooting distance. • Find the coordinates of each corner of the chessboard • Find the coordinator of each corner of the chessboard at sub-pixel points • Finding intrinsic and extrinsic values through each coordinate
  • 7.
    Sets the numberof angles in and the size of the chessboard pattern as well as the distance between the lens and the shooting distance. • The number of angles in the chessboard pattern • The unit size of each pattern • The distance between the camera lens to be stereotyped • The shooting distance between the stereo camera and the chessboard pattern Chessboard patterns used 1
  • 8.
    Find the coordinatesof each corner of the chessboard The next step is to find the coordinates of each corner point on the chessboard using the cv2.findChessboardCorners function. In using this function 3 supporting arguments are needed : • The source of the image • The number of inner corners in the image • The argument of the operation flags. 2 Find the coordinator of each corner of the chessboard at sub-pixel points In the calibration process, accuracy and precision are the most decisive values. To obtain an accurate result, it is necessary to obtain the location of the angle with an accuracy level of up to sub-pixels. 3
  • 9.
    Finding intrinsic andextrinsic values through each coordinate The final step in the calibration process is to pass the 3D point on the real-world coordinates and the 2D location of each coordinate in all images to the cv2.calibrateCamera function. 4
  • 10.
    Result RMSE value withchessboard with a pattern unit size of 3 cm RMSE value with chessboard with a pattern unit size of 11 cm In building a stereo vision system, the RMSE value is output by the cv2 function. Calibrate stereo becomes an evaluative foundation in the calibration process. In this study, two types of chessboard patterns were used with several shooting scenarios for camera calibration needs. Table 1 and Table 2 show the RMSE values obtained from each shooting scenario for calibration needs
  • 11.

Editor's Notes

  • #2 The development of Indonesia's transportation sector is very high, but it needs to be followed by adequate infrastructure development. Based on data from the Central Statistics Agency (BPS) in 2016-2021, the number of vehicles has increased significantly yearly. So, there is often a buildup of vehicles which often results in vehicle accidents [1]. Intelligent Transportation System (ITS) is a vastly developing field in the world of transportation. Automatic detection of vehicle speed and acceleration is part of visual computing in ITS.
  • #3 1. The research conducted by Abdelmoghit Zarane et al. with the title Distance measurement system for autonomous vehicles using stereo camera followed by measuring real-time distance by focusing on the use of stereo cameras, namely two cameras installed in the same horizontal position and moved vertically with a predetermined distance with using stereo dashboard camera. 2. The research conducted by Yasir Dawood Salman et al. with the title of Distance Measurement For Self-Driving Cars Using Stereo Camera This research proposes measuring the distance of objects in self-driving cars designed only relying on stereo cameras good results.