Road Segmentation Using Variant Convolutional Neural Network Architectures Based on Discrete Wavelet Transform

Remote sensing semantic segmentation Satellite images with high resolution have been used in a variety of applications and technologies, including forest fires, storms, traffic management, road extraction, and so on. Roads can be extracted from remote sensing imagery using a variety of models that employ both conventional and deep learning techniques. The conventional methods had a high level of false detection of road extraction, whereas deep learning methods have higher accuracy by obtaining more information. In this study, the use of convolutional neural network models produces precise results when compared to other deep learning and traditional approaches. Road segmentation is used to find damages in roads caused by hurricane-like storms that cause extensive damage to transportation areas, as well as analyzing traffic areas, mapping features, and other factors.