International Journal of Industrial Engineering and Management

Lorem ipsum dolor sit amet, consectetur adipisicing elit, sed do eiusmod tempor incididunt ut ero labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco.

GUIDE FOR AUTHORS SUBMIT MANUSCRIPT
Vol. 14 No. 1 (2023)
Original Research Article

Deep Learning Approach for Volume Estimation in Earthmoving Operation

Faria Alam
Department of Industrial Engineering, Southern Illinois University Edwardsville, Edwardsville, IL, USA
Hoo Sang Ko
Department of Industrial Engineering, Southern Illinois University Edwardsville, Edwardsville, IL, USA
H. Felix Lee
Department of Industrial Engineering, Southern Illinois University Edwardsville, Edwardsville, IL, USA
Chenxi Yuan
Department of Construction Management, Southern Illinois University Edwardsville, Edwardsville, IL, USA

Published 2023-03-30

abstract views: 27 // FULL TEXT ARTICLE (PDF): 0


Keywords

  • Convolutional Neural Network,
  • Deep Learning,
  • Earthmoving Operation,
  • Transfer Learning,
  • Volume Estimation

How to Cite

Alam, F., Sang Ko, H., Lee, H. F., & Yuan, C. (2023). Deep Learning Approach for Volume Estimation in Earthmoving Operation. International Journal of Industrial Engineering and Management, 14(1), 41–50. https://doi.org/10.24867/IJIEM-2023-1-323

Abstract

Earthmoving is a significant activity in most heavy structural designing projects. Earthmoving volume is typically assessed by counting the number of stacked trucks and weighing them on a scale; however, these strategies are error-prone and costly. To address the challenge, this study investigated a deep learning approach for estimating earth volume from photo images of loaded trucks. First, a basic classification model with one convolutional layer has been developed to estimate earth volume by classifying the images into different levels. Next, we applied transfer learning to a pre-trained deep convolutional neural network in order to improve classification performance. For evaluation of the approach, the models have been trained and tested by using images of miniature trucks loaded with different amounts of earth, ranging between 0 and 1000 ml up to six classes at 200 ml intervals. The experimental results showed that the pre-trained network with transfer learning achieved more than 90% accuracy in most cases. The results indicate that the proposed approach has the potential in estimating earth volume in trucks in real-time with minimal intervention by taking images.

 

Article history: Received (August 5, 2022); Revised (February 25, 2023); Accepted (March 1, 2023); Published online (March 8, 2023)

 

PlumX Metrics

Dimensions Citation Metrics