Backfill Quality Monitoring Method of Complex Geological Substation Based on Deep Learning and Edge Computing

Abstract

Earthwork backfilling is an important link in substation construction. In order to improve the backfill quality during construction, this paper introduces the edge calculation framework into the earthwork backfilling project of substation, and realizes automatic monitoring and analysis of backfill quality at the construction site. In view of the shortcomings of time-consuming and labor-intensive manual monitoring, this paper adopts deep learning method to identify the driving track of engineering vehicles and automatically monitor the compaction times of engineering vehicles during backfilling. Combining the convolutional neural networks (CNN) with the region proposal network (RPN), the region where the engineering vehicle is located in the video is extracted, and then the gradient amplitude image for identifying the engineering vehicle is calculated and generated by using the HOG feature. By analyzing the video frame sequence one by one, the driving track of the engineering vehicle can be obtained, and the automatic monitoring of backfill quality can be realized.