Application of Multimedia Information Processing in the Prediction of Geotechnical Parameters

Abstract

Key performance parameters of geotechnical materials significantly impact engineering design and construction. To address challenges in measuring certain parameters, this study proposed a prediction method based on multimedia information processing and deep learning. Acoustic emission and computed tomography scan data was processed to extract features related to Poisson's ratio, the void ratio, the density, and the compression modulus of peat soil. A gated recurrent unit neural network optimized by the particle swarm optimization algorithm was employed for parameter prediction. The results showed that the “particle swarm optimization-gated recurrent unit” model effectively predicted these parameters, with the best performance in predicting the compression modulus and the weakest for the void ratio. This approach provides a novel and reliable method for acquiring and verifying geotechnical parameters.