Internet of Things (IoT) becomes indispensable for transport and automotive industry to advance functions in on-road traffic monitoring. Indeed, smart management tools and machine learning concepts are inevitable in vehicle categorization systems. However, existing systems are only built on individual platforms, and for a most part, the classification accuracy remains limited. In this work, these challenges are tackled by designing a novel convolutional neural network (CNN) that substantially improves on-road vehicle classification. In particular, we experimentally harness, to the best of our knowledge for the first time, two different datasets from separated technological platforms based on close-circuit television (CCTV) and fiber Bragg grating (FBG) sensors, respectively. Standard neural networks in single FBG platform yield limited classification accuracy of only 34% - 62% for <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">AlexNet</i> , 51% - 77% for <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">GoogleNet</i> , 57% - 78% for <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ResNet-50</i> , and 59% - 86% for <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ResNet-101</i> . In contrast, hybrid CNN classification with individual CCTV and FBG datasets substantially improves detection levels, reaching in-class accuracy up to 90% - 97%. Moreover, this classification concept includes an intrinsic back-up verification with respect to each platform compensating the shortcomings of individual technologies. Our demonstration can make key advances towards near-unity accuracy in vehicle classifications for IoT systems, capitalizing on cost-effective and well-established platforms.
Read full abstract