Industrial Part Counting using Computer Vision

Sammanfattning

Reliable quantity verification is important in logistics and manufacturing, whereincorrect deliveries can lead to delays, customer claims, and increased costs. At thecomplete knock down (CKD) operations at Volvo Trucks Tuve, quantity verificationis currently performed manually, making the process time-consuming and vulnerableto human error. Therefore, this thesis investigates the potential of using computervision for industrial part counting.The aim of the thesis was to evaluate computer vision-based part counting at CKDpacking stations by reviewing the current state of the art and assessing the performanceof the Volvo Vision System (VVS). The study followed a design scienceresearch methodology approach and included a current state analysis, stakeholderanalysis, interviews, observations, a literature review, and practical tests using VVSand YOLOv5 object detection models. Different hardware settings, lighting conditions,and dataset sizes were tested to evaluate the feasibility of the system in anindustrial environment.The results show that computer vision-based object counting has potential for industrialapplications, primary in structured environments. The experiments demonstratedthat VVS could successfully detect and count several industrial parts undercontrolled conditions. Increasing the number of training images per class slightlyimproved the model performance, and the system achieved high accuracy for multipleobject classes. However, the study also identified several challenges related tolighting variations, object overlap, reflections, scalability, and long-term maintainability.The study concludes that VVS has high potential and performance when it comes toobject detection. However, in cases such as this one where there is a large amount ofhighly varied products and manual work, there are a lot of challenges that hindersa smooth implementation of a vision system.