Search or add a thesis

Advanced Search (Beta)
Home > Automated scratch detection syseem for the automative industry.

Automated scratch detection syseem for the automative industry.

Thesis Info

Author

Syed Muhammad Hassan Naqvi, Muhammad Ashir Wahid

Program

BS

Institute

Habib University

Institute Type

Private

City

Karachi

Province

Sindh

Country

Pakistan

Thesis Completing Year

2018

Thesis Completion Status

Completed

Subject

Software Engineering

Language

English

Added

2021-02-17 19:49:13

Modified

2024-03-24 20:25:49

ARI ID

1676724375226

Similar


Following the manufacturing process within an automotive industry, a vehicle is prone to defects being present on its body. In the Pakistani industry, one of the largest contributors to these defects are scratches. The current measure being taken to detect scratches on vehicles involve the presence of trained specialists who inspect the body under specific lighting conditions. For this very reason, a low cost and effective automated system is desired which is capable of detection scratches on the front bumper of vehicles on the production line of industries. The main features of this system should its ability to pinpoint a range of areas where the scratches originate from in the manufacturing process as well as to help improve the current manual inspection system in place. To tackle the problem at hand, we propose a scratch detection device, based on a microcontroller interfaced with a distance sensor and a camera module, set up on a robotic arm which is mapped to the front bumper of a vehicle. The system is placed on the manufacturing line. It detects the presence of a vehicle through the threshold being broken on the distance sensor. It then begins the mapping process and takes images of the bumper which it then sends to the image processor located on the microcontroller of the device. The image processor loads the images and begins to preprocess them. This involves the removal of unwanted objects, the removal of reflections of surfaces, etc. It then passes the preprocessed image onto a trained machine learning model which outputs a decision stating whether a scratch is present or not. This data is then logged onto a spreadsheet online which is accessible by the human inspector. The trained machine learning model is based on a Convolutional Neural Network that is trained on a image dataset of around 1500 images of scratches and non-scratches. This model has an accuracy rate of 90% and is capable of telling whether a scratch is present in an image or not. The Robotic Arm has 6 degrees of freedom, allowing it to move in all four directions in a spacial plane as well as extend further into space. This is done to take into account the curved body of a front bumper. The proposed implementation is to place two such Automated Scratch Detection Systems on two sides of the vehicle at a distance of 16 centimeters from the body. This ensures optimal processing time as well as detection accuracy. Each of the systems map towards one half of the bumper and perform the image processing task in a total of 12 minutes
Loading...
Loading...

Similar News

Loading...

Similar Articles

Loading...

Similar Article Headings

Loading...