Optimized deep networks for road extraction using satellite images

Nowadays, satellite images are improved rapidly in digital presentations. It is mainly utilized for feature analysis in different applications. Moreover, it has piqued researchers’ interest in road extraction applications. Several works of literature were implemented in the past based on the neural scheme and optimization features because of the complex data. Hence, a novel Antlion-based Tuned Deep Network (ALbTDN) was implemented with essential characteristics. They apply the trained ALbTDN model to unseen satellite images for road extraction. At first, the collected satellite image data was trained to the system, and the loud characteristics were removed in the pre-processing phase. Then, the refined categorization layer imports data for the feature extraction process. Subsequently, the antlion fitness is used to tune the feature extraction variable after extracting the road features based on the derived features’ intensity classification. Later, the extracted features were incorporated into the antlion fitness, which gave better-forecasted results. The proposed method is used to extract roads from satellite imagery accurately. ALbTDN combines the power of deep learning networks with the optimization capabilities of the Antlion algorithm to enhance the road extraction process. Finally, the scheme was implemented through the MATLAB platform. Moreover, the performance metrics were calculated and equated with the other current models. The present suggested model has recorded the highest read extraction performance outcome. Experimental evaluations on benchmark datasets with an effective ALbTDN algorithm, achieving state-of-the-art results in road extraction accuracy and generalization.