Quantum-enhanced feature extraction for smart logistics and transportation systems
Akshat Bokdia, Nitin Lodha, V. Vijayarajan · 2025
In feature extraction, effective methodology takes prime importance since most algorithms will only be as good as the feature extraction. This chapter discusses three different methods of feature extraction against the German Traffic Sign Recognition Benchmark dataset, an important benchmark for traffic sign image classification. These include Autoencoders, a combination of DWT and SVD, and a Sobel operator paired with DWT and SVD in a hybrid methodology. While Autoencoders derive hierarchical features from the images effectively, Sobel operators paired with DWT and SVD are helpful in detecting edges and capturing multi-resolution details. On the other hand, DWT with SVD targets the extraction of key features from the dataset; hence, an improved approach toward the accuracy and robustness of traffic sign classification, contributed by both, is compared. The chapter also explores quantum approaches, such as Quantum PCA and Quantum Clustering, that improve the effectiveness of these techniques by making feature extraction on large datasets faster, more scalable, and more resilient. The results from the comparative analysis exist in the expectation that such methods would outperform the classical methods both in speed and accuracy. Some of the wide-ranging applications of quantum computing in the logistics and supply chain management area include route optimization, inventory oversight, predictive maintenance, and the multiple transformative capabilities available in these fields through quantum computing. The chapter concludes with a discussion of current constraints as well as potential opportunities to apply quantum computing in current practical logistics frameworks which would enable further breakthroughs in automated traffic systems and intelligent logistics.