The Fine Design Strategy of Urban Streets Using Deep Learning With the Assistance of the Internet of Things

Abstract

In order to achieve the fine design of urban streets in the Internet of Things (IoT) environment, this work proposes a semantic segmentation network based on improved DeepLabV3+ to perform semantic segmentation of street scenes. Considering the processing of images’ most basic edge features, this work proposes a street view image’s enhancement algorithm based on the Generative Adversarial Network (GAN). This algorithm utilizes GAN’s generator network to generate high-quality street view images and evaluates their authenticity using a discriminator network. Meanwhile, this algorithm also introduces an adaptive weight control method to optimize the interaction between the generator and discriminator networks. The simulation experiment is conducted. The results show that the improved GAN algorithm outperforms other algorithms in various reconstructed image indicators. Its PSNR values on the A, B, and C datasets are 37.432, 35.925, and 37.127, respectively. It reveals that the improved GAN algorithm based on edge detection can achieve high restoration results in the application of super-resolution reconstruction of street view images, which has a certain reference value for the fine design of urban streets.

Keywords

• Internet of Things
• deep learning
• urban streets
• fine design
• generative adversarial network