PEPC

PEPC is a libary that contains several image processing applications, and my Bachelor Final Project.

Table of Contents

1. About The Project
   • Built With
   • Prerequisites
2. Project Details Steps
   • Development Steps
   • Usage
3. Acknowledgments

About The Project

PEPC has 4 different capabilities. These are

• Pattern Evaluation
• Pattern Matching
• Natural Feature Marker Pose Calculation 6 DoF
• Cross Correlation Calculation using Template Matching

(back to top)

Built With

• C++ (ISO C++14 Standard)
• OpenCV
• Tesseract OCR

(back to top)

Prerequisites

Make sure you fulfilled the prerequisities before using the library.

• OpenCV 4.5.3
• C++ (ISO C++ 14 Standard)
• Tesseract 4.1

Project Details

Development Steps

In this section I listed development stages of the PEPC library.

1. I made lots of research about template matching, pattern matching algorithms.
2. I implemented OpenCV's pattern matching tutorials before I customized them for my purpose.
3. I developed pattern matching algorithm using OpenCV's documentations and tutorials.
4. I made researches about pose calculation and studied fundementals of image processing using the course Introduction to Computer Vision.
5. Later on I developed natural feature marker pose calculation algorithm in light of previous researches.
6. Made reserach about pattern evaluator systems, I read various articles and tried to implement some approaches mentioned on those articles.
7. I tried to add more features to the pattern evaluator algorithm.
8. I worked on cross correlation and created an algorithm to determine a switch's state, and number written on that switch.

(back to top)

Usage

Let's talk what this library capable of!

• Pattern Evaluator: This brings ability to check few properties of the image target, which intended to be used as a natural marker. By checking properties that are likely to effect tracking quality PEPC may give feedback about the quality of the pattern. These properties may listed as, Checking,

  • brightness of the environment
  • total keypoints found in the image
  • total inliers number
  • selected pattern to frame size ratio

• Matching Algorithm: This algorithm is developed using ORB keypoint extractor, BRIEF descriptor, MAGSAC++ estimator. It tracks pattern by matching keypoints from the captured frame during run time.

• Pose Calculation: After having PEPC's pose calculation algortihm, you will be able to use naturaş feature markers to calculate camera pose in 6 dof.

• Finding Dissimilarity: It is possible to determine two different states of a product (such as switch open or switch close) by using OpenCV's template matching alogrithm. It does provide cross correlation of given two images. Here I used tesseract library to read numbers on switch and tried to use them as their ID.

• I build tesseract using vcpkg. vcpkg install tesseract:x64-windows-static

Acknowledgements

I would like to thank to my supervisor Resul Aydoğan and Sefa Ödemiş for all their help and advice with this work. I would also like to thank rest of the HARF team for their support during my co-op journey.