AUTOMATED QUALITY CONTROL OF BISCUITS USING COMPUTER VISION

Abstract

This thesis is intended to describe the design and fabrication of a solution for Automated Quality Control of the Biscuits through Computer Vision. The project is aimed at providing an optimized solution for the quality control of biscuits which can be deployed at local industries in order to automate the process. The goal was to design, fabricate, test, and validate a working prototype of the solution. Our target was to provide a well-integrated automated solution for the industries in order to localize this technology, to reduce human interaction in the production line and to ensure hygiene of the eatables. We also aim to reduce the labour cost and make the process cost effective. Its scope can be expanded to industries where quality is currently ensured via human vision. To materialize a machine with such capabilities, we had to work in parallel to take right decisions regarding the hardware involved, and the software algorithm driving it. The complete setup should be able to differentiate between standard and sub-standard biscuits along with a responsive actuation mechanism driven by microprocessors. Two prototypes were developed to achieve the set objective for the project. The first prototype V1 was to be developed as a proof of concept, and the insights gained from the first attempt assisted us in fabricating a much efficient, practical, and robust final prototype V2. The objectives were successfully achieved with some minor shortcoming in the final prototype. Further work can be done on it to improve the colour resolution, robustness, and serviceability. Machine learning can also be introduced in the algorithm for enhanced capabilities.