Coloured image classification with quantum machine learning algorithms for intelligent transportation systems

Coloured image classification with quantum machine learning algorithms for intelligent transportation systems

Quantum computers have great potential to change the future of Artificial Intelligence (AI) applications for practical problems. There currently exist some algorithm design that takes exponential time on a classical computer can take polynomial time on a quantum computer. With high demand for fast and reliable AI applications, such as traffic sign recognition for Intelligent Transportation Systems, it is beneficial for society and smart infrastructure to develop and utilise the most suitable AI algorithms for Quantum Computers (QCs). This Ph.D research project aims to focus on the feasibility of QCs to implement the image multi class classification techniques used in various domains, including AI applications. Moreover, in this study new successful techniques such Quantum Neural Network algorithms will also be explored to improve the efficiency of quantum image processing algorithms on QCs based on the similarity of these techniques with the mechanics of the QCs. This PhD research is expected to contribute to the knowledge domains in areas of real-life road traffic signs. The objectives of the research are: (i) To develop a new image-based multiclass classification algorithm using the quantum entanglement approach in comparison with classical machine learning (CML). (ii) To design a new filter for image processing to determine efficiency of quantum machine learning over CML. (iii) To investigate proposed quantum filter for binary image classification of complex traffic signs on small sample data. Within the context of these goals, this project is a partnership between Commonwealth Scientific and Industrial Research Organization (CSIRO) and University of Southern Queensland, which aims to build cutting-edge algorithms and novel methods for future QCs. The designed algorithms will be implemented in the Python software interface on available online quantum simulators. The scientific contributions indicate remarkable improvement in our model's performance on black and white images. However, we did not observe Quantum Machine Learning (QML) performing better for complex traffic signs using multi-class classification. Nevertheless, we achieved good results with binary image classification. This research will contribute to the scientific foundation for future applications of AI in quantum computers.

File reproduced in accordance with the copyright policy of the publisher/author.