Corridor Mapping Processing Using the Machine Learning Approach

The study investigates the use of machine learning in remote sensing to identify and map linear features such as roads, pipelines, and utilities in civil engineering. Remote sensing involves using sensors to gather data about the Earth's surface and atmosphere from a distance. Machine learning is a powerful tool that can be used to analyze and interpret this data to extract meaningful insights and make predictions. The availability of large amounts of data from remote sensing technologies has greatly increased in recent years. It has become increasingly more work to analyze and interpret this data manually. Machine learning solves this problem by automating the analysis and interpretation of remote sensing data. Using machine learning in remote sensing can provide numerous benefits, including increased efficiency and accuracy in data analysis. In this study's context, the primary dataset's accuracy, which only contained RGB values, was 83%. The accuracy was improved to 89.9% when an integrated dataset containing RGB and elevation information was used. The study also compared pixel-based and object-based classification using a random forest algorithm and found that object-based classification led to slightly improved accuracy. Furthermore, the accuracy was improved from 89.9% to 91.11% when using a deep learning convolutional neural network (CNN), even for a tenfold larger area. The training time for the CNN algorithm model was 6x longer than the traditional machine learning model. However, implementing the trained model over large areas took only minutes rather than hours and enabled the extraction of single types of features without any redundant data. Similar results were obtained in a second example applied to a road corridor. The findings of this study can have practical applications, such as the analysis of infrastructure safety and digital twins. Machine learning can analyze data from remote sensing over time to detect changes and trends, which can be useful for understanding environmental impacts and identifying potential conflicts. Additionally, the results of this study can inform the design and planning of urban corridors with data credibility in mind by helping to identify the potential hazard and possible solutions