A Machine Vision System for Real-Time Can Counting Group 52 - ENGR 499 Detailed Design Report Team Captain - Software Architecture - Matthew Tucsok 37924164 Data Science - Todd Charter 54151162 Coordination - Jack McClelland 31046162 Mount Design - Mihir Doshi 45622164 Mount Design - Zachary Carels 51275155 Faculty Advisor: Dr. Homayoun Najjaran Industry Partner: Jay Aarsen UBC OKANAGAN School of Engineering University of British Columbia - Okanagan Date Submitted: 2021-04-13 Submitted to Dr. Homayoun Najjaran
Executive Summary In the recycling industry, bottle depots want to maximize the value of their aluminum bales. To determine the price of a bale, it is either measured by the weight of the bale or by the number of aluminum cans in the bale. Due to the nature of these aluminum cans, liquids may still be present in the containers which causes an overestimate in the weight of the total aluminum content. As a result, pricing is generally lower when based on weight as opposed to a total can count, making it more profitable to sell the scrap aluminium by the can. Building off a prior project for Interior Freight and Bottling Depot and in collaboration with Okanagan College, a machine vision system is designed that is able to count aluminum cans with greater than 95% accuracy as they are being crushed and baled. This accuracy is achieved under ideal conditions. Additionally, real-time counting is achieved with sufficiently powerful computing hardware. Previous implementations of this project have explored physical and theoretical solutions with varying levels of success. One implementation used a shaker table to organize the cans and count them with photo-diodes. Although this was unsuccessful, it led to the idea of aiming a camera at the conveyor to implement a machine vision counting system. Primary stakeholders include: Jay Aarsen, CEO and designated industry partner from In- terior Freight and Bottling Depot; Yves Gagnon and Dr. Andrew Hay, who represent the engineering technology departments at Okanagan College; and Dr. Homayoun Najjaran, whose lab specializes in machine vision applications for automated systems. i
Contents 1 Problem Specification 1 1.1 Background Information 1 1.2 Problem Statement 2 2 Needs and Constraints Identification 3 2.1 Stakeholders . 3 2.1.1 University of British Columbia 3 2.1.2 Okanagan College . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 2.1.3 Interior Freight Bottle Depot 3 2.2 Constraints 4 2.3 Considerations . . . . . . 2.3.1 Economic Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2.3.2 Environmental Considerations . . . . . . . . . . . . . . .