Change Detection of Grid Construction Based on Satellite Remote Sensing Images and Siamese Neural Network

Abstract

Quality monitoring of grid construction process based on satellite remote sensing method can standardize the quality behavior of all parties involved in the engineering construction and related institutions and evaluate the environmental and safety impacts of the grid engineering. Thus, a Siamese-Neural-Network-based approach using 3-channel remote sensing images was proposed to quickly monitor grid construction changes. This approach can localize the changed buildings with the help of previous and post remote sensing images of the interesting regions, classify the changes into increased buildings and decreased buildings, and output the masks reaching pixel-level positioning accuracy. The proposed method was tested and compared with other methods based on 2-channel. It has higher accuracy and better robustness and is suitable for the above-mentioned scenes of changes in grid construction. Through experimental research, it is proved that the use of multi-temporal high-resolution remote sensing data and deep learning methods can not only monitor the overall construction status and progress of the project in a timely manner, realize the monitoring of large-scale passages or site clearance, but also monitor the construction scope in a timely and effective manner. The dynamic changes of the ecological environment and the review and evaluation of the construction workload.