OpenCV Projects (C++ | CMake)

A curated collection of foundational computer vision projects built using OpenCV (C++) and CMake, focused on real-time processing, classical vision pipelines, and correct project structuring.

These projects are intentionally designed to:

• Emphasize core OpenCV concepts

• Avoid black-box abstractions

• Be buildable, debuggable, and extensible

• Help beginners understand why things work, not just that they work

📁 Project Overview OpenCV_Projects/ │ ├── projects/ │ ├── Color_Picker/ │ ├── Virtual_Painter/ │ ├── Documents_Scanner/ │ └── Russian_Number_Plate_Detector/ │ └── README.md

Each project is self-contained and follows the same structure:

<Project_Name>/ ├── CMakeLists.txt ├── src/ │ └── main.cpp └── Resources/ (only when required)

🔧 Build Requirements

• C++17 or later

• OpenCV ≥ 4.x

• CMake ≥ 3.16

• Windows / Linux (tested on Windows with MSVC)

Make sure OpenCV binaries are available in your system PATH (Windows users).

🟦 1. Color Picker Purpose

Interactively identify HSV color ranges from a live camera feed using trackbars.

Key Concepts

• BGR → HSV color space conversion

• Real-time trackbars

• Binary masking using inRange

• Practical color calibration workflow

How It Works

• Camera feed is converted to HSV

• Trackbars control H, S, and V min/max values

• Masked output updates live

• Final HSV ranges are used by other projects (e.g. Virtual Painter)

Why This Project Matters

Color calibration is a prerequisite for many vision tasks. This project teaches how to do it correctly instead of guessing values.

🖌️ 2. Virtual Painter Purpose

Draw on a virtual canvas using a colored pen/marker tracked in real time.

Key Concepts

• HSV-based object tracking

• Contour detection

• Centroid computation using moments

• Persistent drawing using an overlay canvas

How It Works

1. HSV ranges (from Color Picker) isolate the pen color

2. Largest valid contour is selected

3. Contour center is tracked frame-to-frame

4. Lines are drawn on a separate canvas and blended with the frame

Design Notes

• One canvas per session (no flickering)

• Noise filtered via contour area threshold

• Explicit handling of lost tracking frames

Limitations (Intentionally Left Visible)

• HSV sensitivity to lighting

• No Kalman filtering (kept simple on purpose)

📄 3. Document Scanner Purpose

Simulate a mobile document scanning pipeline using classical image processing.

Key Concepts

• Edge detection (Canny)

• Morphological operations

• Contour approximation

• Perspective transformation

• ROI cropping

Pipeline

1. Convert image to grayscale

2. Blur and detect edges

3. Dilate edges to strengthen contours

4. Detect largest quadrilateral contour

5. Apply perspective warp

6. Crop borders for clean output

Why This Is Important

This project demonstrates how geometry + image processing can solve real-world problems without deep learning.

Practical Notes

• Resource paths are resolved at runtime

• Designed to avoid hard-coded absolute paths

• Easy to adapt for camera input later

🚘 4. Russian Number Plate Detector Purpose

Detect Russian vehicle number plates in real time using a webcam.

Key Concepts

• Haar Cascade classifiers

• XML-based object detection

• Real-time video processing

• Robust resource loading

How It Works

• Webcam (VideoCapture(0)) provides frames

• Haar cascade XML is loaded at runtime

• Plates are detected using detectMultiScale

• Bounding boxes are drawn live

Important Design Choice

This project uses classical Haar cascades deliberately:

• Lightweight

• No GPU required

• Easy to understand

• Suitable for embedded or low-resource systems

Known Limitations

• Cascade performance depends on training data

• Sensitive to extreme angles and lighting

• Detection only (no OCR yet)