Abstract
Internet of things (IoT) primarily aims to realize valuable services such as smart homes, smart buildings, and smart transport. To this end, smart applications are faced with different challenges, particularly in the implementation of a smart transportation system, which requires maximum road and travel safety. Road crack detection has been extensively studied and presented with various solutions; however, these approaches are limited by the inhomogeneity in the crack intensity and background complexity, such as a shadow with similar intensity and pavement contrast, which are known obstacles in the accurate prediction of road cracks. To overcome these issues, an IoT system with a bio-inspired deep learning approach was introduced herein for accurate road crack detection. In the proposed approach, transportation images are first collected using a smart mobile sensor, then processed by a bio-inspired self-learning co-evolutionary deep-convolution neural network. The optimized neural networks provide the required framework by analyzing the collected images to detect cracks more accurately. The efficiency of the proposed system was confirmed in different metrics, including the per-pixel accuracy (99.04%), Jaccard index (98.42%), loss error rate (0.03), precision (99.25%), recall (99.24%), and prediction accuracy (99.72%) metrics.