Object Detection

Project Introduction

Machine Learning (ML) technology has been developing for decades. One of the applications of ML is object detection. Object detection is a new technology that uses a series of technologies including computer vision and image processing to detect instances of objects of a particular class, such as people, dog and phone. One of the applications of object detection is in chemical industry. Usually, the chemical matters are carried by barrels using particular materials and most of them are toxic. To manage these barrels, lots of labors have to be deployed in the field. Thus, in order to reduce the labor used in the field and increase the safety in such an environment, a tracking system for these barrels are necessary.

Provided Solution

This project is to develop an object detection submodule in a tracking system. This entire tracking system includes object detection submodule, front-end service software, interfaces and hardware components. By using such a system, the staff are able to know the situation of every key node, and able to search and obtain the information of these chemical barrels. In this project, I designed and developed this submodule using YOLOv3 pretrained model. By using vast number of images as a database to train the pre-trained model on cloud using GPUs, I am able to obtain the trained model and integrate the model into my program using C# in .NET Windows Application. after successfully detecting the object in the video stream from IP cameras through Onvif protocol, the application will automatically read the QR codes that are attached to the barrels. The information inside the QR codes will be then sent to the central system to further process. Finally, the service submodule will process the information and display them on the user interface. As a result, this submodule is embedded in a real system using Windows based server for a chemical factory.