🔍 Weld Continuity Check

An automated computer vision script to detect imperfections in T-fillet welds. This project uses image processing techniques to distinguish between continuous weld beads and discontinuous defects like porosity or cracks.

🛠️ Tech Stack

• Language: Python 3
• Library: OpenCV (cv2), NumPy
• Technique: Canny Edge Detection, Morphological Filtering

✨ Features

• Region of Interest (ROI): Automatically masks the top and bottom of the image to focus strictly on the welding zone.
• Noise Reduction: Uses Gaussian Blur to filter out micro-spatter and surface texture.
• Smart Filtering: Distinguishes defects based on contour length:
  • Ignored: Small noise (< 10px) and continuous weld lines (> 100px).
  • Detected: Medium-sized broken segments (10px - 100px) representing defects.
• Coordinate Logging: Outputs exact (x, y) coordinates of every detected imperfection.

🚀 How It Works

1. Preprocessing: The image is converted to grayscale and smoothed with a 5x5 Gaussian Blur.
2. Masking: A rectangular mask creates a "tunnel vision" effect, isolating the center weld.
3. Edge Detection: cv2.Canny (Thresholds 50, 150) identifies high-contrast edges.
4. Analysis: The script calculates the arcLength of every contour. Continuous lines are marked green (safe), while broken lines are marked red (defects).

💻 How to Run

1. Clone the repository.
2. Ensure you have the required libraries:

   pip install opencv-python numpy

3. Place your weld image (e.g., badweld2.jpg) in the project folder.
4. Run the script:

   python weld_check.py

🧠 What I Learned

• How to use Canny Edge Detection to extract structural features from metal surfaces.
• Using Bitwise Operations (Masking) to focus algorithms on specific image regions.
• Calculating Moments to find the centroid coordinates of irregular shapes.

weldcheckvideo.mp4