1 of 12

Automated Industrial Inspection System

Team Members:

​

​

​

​

​

  1. Anthony Nikhil Reddy Lingala

2.    Arsalan Anwar

3.    Shaktidhar Kanamarla

4.    Kurapati Aravind​

2 of 12

Agenda

�

​

������

  • Overview
  • Features and Techniques
  • Demo
  • Results
  • Challenges and Limitations
  • Future Steps

This Photo by Unknown author is licensed under CC BY-SA-NC.

3 of 12

Overview

In today's manufacturing environment, maintaining high product quality is crucial. Manual defect detection is time-consuming, prone to errors, and inconsistent. 

Automated defect detection systems leverage computer vision techniques to check if the product flow is right, identify defects and deformities in products, significantly improving quality control processes.

This Photo by Unknown author is licensed under CC BY-SA-NC.

4 of 12

Features 

Automated Industrial Inspection System consists of three main features:

  • Deformity Checker: This checks if there are any deformities in the shape of the product
  • Product Packaging Checker: This checks if the right number of products are being filled and packaged in the carton boxes
  • Defect Detector: This scans and identifies if any defects are present in the product

Deformity Checker

Product Packaging Checker 

Defect Detection 

5 of 12

Deformity Checker

    • Identify deformities in the product using object boundaries for verifying part dimensions and shapes.

Purpose:

    • Extracted focus area from the video
    • Utilized Thresholding and Edge detector to highlight areas with significant changes in image intensity.
    • Denoised the bottle frame (erode and dilate)
    • Non-focus Area could be blurred or left as it is for performance improvements

Technique Used:

Deformity Checker

6 of 12

Product Packaging Checker 

    • This checks if the right number of products are being filled and packaged in the carton boxes
    • Helps avoid void packaging, a common problem where empty/no product is packaged

Purpose:

    • Utilized an ensemble of image segment extractor, template matching and match redundancy remover to identify, trace and count the products on a specific region of the conveyor belt

Technique Used:

Product Packaging Checker 

7 of 12

Defect Detector

    • This scans all the images captured for a particular batch and classifies whether the product is defective or not
    • Batch processes any folder that is given as an input

Purpose:

    • Utilized a multi-layer template matching architecture along with feature-based matching to compare parts against standard templates
    • First Layer checks if the product is defective and identifies only the region of defect
    • Second Layer checks if the product is non-defective and highlights the entire product

Technique Used:

Defect Detector

8 of 12

Demo

9 of 12

Results

Benefits of Automation:

  • Reduces health and safety risks
  • Decreases operational costs and liability
  • Improves product quality
  • Helps reduce human bias

​

Economic Advantages:

  • Lowers labour costs
  • Reduces downtime and production delays
  • Fewer product defects and recalls
  • Enhances overall product quality and consistency

​

10 of 12

Challenges & Limitations

Bottles can come in various shapes, sizes, and colours. System must be adaptable for each design

Changes in lighting conditions, dust, and other environmental factors can affect the performance of the system

Integrating new vision systems with existing manufacturing and quality control processes can be complex

 Speeds at which production lines operate require the system be capable of capturing and processing images quickly enough to keep up

11 of 12

Future Steps

 Leverage deep learning models to improve the accuracy and versatility of the system by enabling them to learn from a broader set and adapt to new bottle designs.

 Incorporate 3D imaging techniques to detect volumetric defects and the overall geometrical assessment , which are difficult to gauge using only 2D images.

 IoT devices could be used to collect additional data from the manufacturing environment, such as temperature and humidity,  to adjust the algorithms in real time.

12 of 12

Questions?