Image Analysis for Automated Defect Detection

Project Domain / Category

Image Processing

Abstract / Introduction

In sectors like oil and gas, manufacturing, and civil engineering, ensuring the structural integrity of industrial pipelines and building surfaces is critical for safety, compliance, and cost reduction. Manual inspection methods are labor-intensive, inconsistent, and not scalable for large infrastructures. This project proposes an automated defect detection system using established image processing techniques to assist in identifying visible defects on surfaces such as pipelines (e.g., cracks, dents) or buildings under construction (e.g., wall cracks, surface degradation, concrete spalling).

The system will utilize classical image processing operations such as grayscale conversion, noise filtering, edge detection, thresholding, and contour analysis. Optionally, students may integrate a lightweight pertained models (e.g., YOLOV4-Tiny) for improved accuracy. A basic user login system, image upload interface, and defect visualization panel will be provided for ease of use. The system will support local image storage and simple reporting.

Note for Students: Cloud resources (e.g., Google Colab) or personal computing devices with sufficient processing power.

Functional Requirements:

1.      FR1: User authentication system with login/logout functionality.

2.      FR2: Image input system for capturing product images.

3.      FR3: Image preprocessing (resizing, noise removal, brightness adjustment).

4.      FR4: Integration of pretrained object detection model (e.g., YOLOv4-Tiny) to detect visible defects such as cracks, leaks, or surface irregularities.

5.      FR5: Display of processed images with defects visually highlighted (bounding boxes or labels).

6.      FR6: Display of detected defects with highlighted areas.

7.      FR7: Report generation with defect count, defect type, and marked images.

Tools:

·        Programming Language: Python

·        Frameworks/Libraries: OpenCV, NumPy, SciPy

·        Development Environment: Jupyter Notebook, PyCharm, VS Code

·        Additional Tools: Simple local file storage for image saving

Supervisor:

Name: Dr. Sana Rao

Email ID: sana.rao@vu.edu.pk

Skype ID: rao.sana10