Introduction

This repo contains works of CENG391 course,Introduction to Image Understanding, and my studies about computer vision area.

Laboratory Works

Both C++ and Python's OpenCv library is used during laboratory studies of this course.The content of homeworks listed below.

• OpenCV Installation and Setup
• Basic data types and UI with OpenCV
• Smoothing & Sharpening operations
• Image compositing with Laplace pyramids
• Harris corner detection
• Image stitching for panorama generation
• Measuring the keypoint recognition rate, FAST & BRIEF
• Homography estimation, DLT
• Homography estimation with RANSAC

Homeworks

• Manuel image correspondence
• Bilateral filter algorithm
• Bilinear interpolation
• Object detection with RANSAC
• Epipolar Line

Internship Projects

• project_rollor_deformation_detection

  • This project is about finding deformation on rollers during production process in a factory. Normally, there are four different text on a roller. However, there may be some missing text due to production errors so the aim of this project is to find rollers that has missing text on it.

  

  Figure1: A Roller sample

• project_balata_deformation_detection

  • This project is about finding deformation on rollers during production process in a factory. Normally, a product should has specific dimensions. However, dimensions of a product can change due to some production errors. This project aims to find products which don't have desired dimensions.

    

    Figure2: A sample product

• project_ffmpegCapture

  • This project implements capturing a camera stream using ffmpeg library of C++ language. OpenCv library has a videoCapture library for capturing videos. However, I tried to implement a C++ class to capture frames more efficiently.

• project_count_people

  • This project aims to count number of people entered a shop during the day using camera stream.

• project_create_heatmap

  • This project aims to create heatmap of customers in a market. It creates heatmap for every hour in a day so it can be used to determine the most crowded hours of a day and the most used places in a market.

    

    Figure3: Heatmap of a market using fisheye camera