Research on Post-Earthquake Landslide Extraction Algorithm Based on Improved U-Net Model

Abstract

Seismic landslides are the most common and highly destructive earthquake-triggered geological hazards. They are large in scale and occur simultaneously in many places. Therefore, obtaining landslide information quickly after an earthquake is the key to disaster mitigation and relief. The survey results show that most of the landslide-information extraction methods involve too much manual participation, resulting in a low degree of automation and the inability to provide effective information for earthquake rescue in time. In order to solve the abovementioned problems and improve the efficiency of landslide identification, this paper proposes an automatic landslide identification method named improved U-Net model. The intelligent extraction of post-earthquake landslide information is realized through the automatic extraction of hierarchical features. The main innovations of this paper include the following: (1) On the basis of the three RGB bands, three new bands, DSM, slope, and aspect, with spatial information are added, and the number of feature parameters of the training samples is increased. (2) The U-Net model structure is rebuilt by adding residual learning units during the up-sampling and down-sampling processes, to solve the problem that the traditional U-Net model cannot fully extract the characteristics of the six-channel landslide for its shallow structure. At the end of the paper, the new method is used in Jiuzhaigou County, Sichuan Province, China. The results show that the accuracy of the new method is 91.3%, which is 13.8% higher than the traditional U-Net model. It is proved that the new method is effective and feasible for the automatic extraction of post-earthquake landslides.

Remote Sensing for Assessing Landslides and Associated Hazards

“…Particularly when modelling is based on spectral information only and training samples for supervised methods randomly selected. However, other studies (Lei et al 2019a, b;Liu et al 2020) highlight the performance of FCN to learn better image features to improve landslide inventory and Landslide Susceptibility Mapping (Zhao and Du 2016;Fang et al 2020).…”

Mapping Post-Earthquake Landslide Susceptibility: A U-Net Like Approach

“…The encoding path usually includes many convolutional and max pooling layers while the decoding path includes convolution and upconvolution layers. U-net has been very successful in many semantic segmentation tasks [34][35][36][37][38], but no literatures have reported the application of U-net for LSM. Traditional U-net model architecture [65] (example for 32 × 32 pixels in the lowest resolution and binary classification output).…”

“…Traditional CNN model can mine the spatial structure information for LSM task, but the original influencing factors of each pixel may be weakened during the convolution and pooling operations. So, in this paper, we develop a U-net like model for mapping post-earthquake landslide susceptibility because U-net model connects encoding path and decoding path to compensate for the information eliminated by decoding [35,39]. What we are concerned about is how to balance the influence of a pixel itself, and its surrounding pixels, to perform an optimal LSM task.…”

“…For example, Sameen [32] developed an one-dimensional convolutional network (1D-CNN) for an LSM task in Southern Yangyang Province, South Korea, which showed better performance than ANN and SVM methods; Wang [33] observed that a CNN showed performance equivalent to or better than an optimized SVM for LSM tasks in Yanshan County, China. As one of the famous CNN models, U-net has been very successful in many semantic segmentation tasks [34][35][36][37][38] but no literatures have reported the application of U-net for LSM. A U-net model consists of an encoding path and a decoding path and connects the two paths to compensate for the information eliminated by decoding because fine information may be lost by pooling and deconvolution [35,39].…”

“…As one of the famous CNN models, U-net has been very successful in many semantic segmentation tasks [34][35][36][37][38] but no literatures have reported the application of U-net for LSM. A U-net model consists of an encoding path and a decoding path and connects the two paths to compensate for the information eliminated by decoding because fine information may be lost by pooling and deconvolution [35,39]. This characteristic is important for LSM tasks as the original influencing factors of each pixel may be weakened during the convolution and pooling operations.…”