Object Detection and Tracking for Crate and Bottle Identification in a Bottling Plant Using Deep Learning

Abstract

This paper presents a practical implementation of supervised object detection techniques for real-world manufacturing applications, specifically for crate tracking, bottle counting, and bottle inspection in a bottling plant. The proposed model architecture utilizes a two-stage tracking process, employing a wide-angle camera and advanced object detection algorithms to overcome the limitations of traditional Convolutional Neural Networks (CNNs). The first stage of the model tracks the crates, while the second stage identifies the bottles within the crates. The accuracy of the proposed approach is validated to be over 99.9%. The paper details the dataset preparation, model architecture, training procedure, and evaluation results.

Received: 24 February 2023 | Revised: 30 March 2023 | Accepted: 10 April 2023

Conflicts of Interest

The author declares that he has no conflicts of interest to this work.