DieCenterLib 1.0.0

DieCenterLib

A lightweight, high-precision die center detection library designed for personal or internal use.

⚠️ Note: This library is tailored for specific die images that contain an inner seal ring.
It is not intended as a universal die/wafer detection framework.

Table of Contents

• Features
• Supported Scenarios
• How It Works
• Pipeline Overview
• Installation
• Basic Usage
• API Summary
• Troubleshooting
• Recommended Image Conditions
• License
• Acknowledgements

Features

• Automatic or manual ROI selection
• Gradient-based edge extraction (Scharr filter)
• Canny edge detection with adaptive behavior
• Outlier rejection using k-sigma filtering
• Linear regression for the four die edges
• Intersection-based corner finding
• Subpixel die center estimation
• Residual statistics: RMSE, Std, MaxAbs, R²
• Optional ROI intensity export to CSV

Supported Scenarios

This library works best when:

✔ The die has a clear rectangular boundary
✔ There is a seal ring (inward dark/bright border)
✔ Lighting is relatively even
✔ Edges are not heavily occluded
✔ Input resolution ≥ 800 × 800 px

How It Works

DieCenterLib extracts die boundaries by analyzing gradient magnitudes within predefined band regions near each side of the ROI:

1. ROI Selection

   • Auto ROI using contour detection
   • Or manual ROI defined by user

2. Gray Conversion + Scharr Gradient
   High-accuracy gradient operator for clean edge responses.

3. Band Extraction (70–80% coverage)
   Each edge is analyzed only in its dedicated band
   → avoids mixing with corners or noise.

4. Canny Edge Detection
   Detects candidate edge points.

5. Line Fitting (with Outlier Removal)
   Using OpenCV’s robust fit + k-sigma rejection.

6. Corner Finding
   Intersection of opposite edges.

7. Center Calculation
   Intersection of diagonals formed by TL–BR and TR–BL.

Pipeline Overview

            +-------------------+
            |   Input Bitmap    |
            +---------+---------+
                      |
                      v
            +-------------------+
            | ROI Selection     |
            | Auto / Manual     |
            +---------+---------+
                      |
                      v
            +-------------------+
            | Gray + Scharr     |
            | Gradient (Gx, Gy) |
            +---------+---------+
                      |
                      v
            +-------------------+
            | Region Separation |
            | (Top/Bottom/...)  |
            +---------+---------+
                      |
                      v
            +-------------------+
            |  Canny Edges      |
            +---------+---------+
                      |
                      v
            +-------------------+
            |  Extract Points   |
            |  (per side)       |
            +---------+---------+
                      |
                      v
            +-------------------+
            | Line Fitting      |
            | + Residual Stats  |
            +---------+---------+
                      |
                      v
            +-------------------+
            | Corner Detection  |
            +---------+---------+
                      |
                      v
            +-------------------+
            | Center Position   |
            +-------------------+

Installation

Install-Package DieCenterLib

Dependencies / Prerequisites

DieCenterLib does not include OpenCvSharp4 binaries.
Before using this library, please install the following NuGet packages into your project:

• OpenCvSharp4 (tested with 4.11.0.20250507)
• OpenCvSharp4.runtime.win (same version as OpenCvSharp4)
• OpenCvSharp4.Extensions (same version as OpenCvSharp4)

Example (Package Manager Console):

Install-Package OpenCvSharp4 -Version 4.11.0.20250507
Install-Package OpenCvSharp4.runtime.win -Version 4.11.0.20250507
Install-Package OpenCvSharp4.Extensions -Version 4.11.0.20250507

Basic Usage

using DieCenterLib;

var parameters = new DieCenterParameters
{
    UseAutoRoi = true,
    RoiWidth = 480,
    RoiHeight = 480,
    Coverage = 0.8f,
    BandWidth = 50,
    MinGrad = 5.0f,
    KSigma = 2.0,
};

var result = DieCenterDetector.Run(bitmap, parameters);

// Die center in original image coordinates
var center = result.CenterInImage;

Console.WriteLine($"Center X={center.X}, Y={center.Y}");
Console.WriteLine(result.SummaryText);

API Summary

DieCenterParameters

Troubleshooting

Problem: Edges missing

• Increase BandWidth
• Increase MinGrad
• Ensure ROI is correct

Problem: Weak contrast

• Apply CLAHE
• Increase exposure
• Increase Canny thresholds

Problem: Bad line fitting

• Increase KSigma
• Check extracted points

Problem: Auto ROI incorrect

• Switch to ManualRoi

Recommended Image Conditions

✔ Stable illumination
✔ Visible seal ring
✔ Moderate contrast
✔ No large scratches
✔ 800×800 px or higher resolution

License

• MIT License © 2025 Napat Sutikant
• Uses OpenCvSharp (Apache 2.0).

Acknowledgements

• OpenCvSharp4 by shimat
• OpenCV Community