MRIFE: A Mask-Recovering and Interactive-Feature-Enhancing Semantic Segmentation Network For Relic Landslide Detection

TL;DR

A semantic segmentation model, termed mask-recovering and interactive-feature-enhancing (MRIFE), is proposed for more efficient feature extraction and separation and greatly improves the performance of relic landslide detection.

Abstract

Relic landslide, formed over a long period, possess the potential for reactivation, making them a hazardous geological phenomenon. While reliable relic landslide detection benefits the effective monitoring and prevention of landslide disaster, semantic segmentation using high-resolution remote sensing images for relic landslides faces many challenges, including the object visual blur problem, due to the changes of appearance caused by prolonged natural evolution and human activities, and the small-sized dataset problem, due to difficulty in recognizing and labelling the samples. To address these challenges, a semantic segmentation model, termed mask-recovering and interactive-feature-enhancing (MRIFE), is proposed for more efficient feature extraction and separation. Specifically, a contrastive learning and mask reconstruction method with locally significant feature enhancement is proposed to improve the ability to distinguish between the target and background and represent landslide semantic features. Meanwhile, a dual-branch interactive feature enhancement architecture is used to enrich the extracted features and address the issue of visual ambiguity. Self-distillation learning is introduced to leverage the feature diversity both within and between samples for contrastive learning, improving sample utilization, accelerating model convergence, and effectively addressing the problem of the small-sized dataset. The proposed MRIFE is evaluated on a real relic landslide dataset, and experimental results show that it greatly improves the performance of relic landslide detection. For the semantic segmentation task, compared to the baseline, the precision increases from 0.4226 to 0.5347, the mean intersection over union (IoU) increases from 0.6405 to 0.6680, the landslide IoU increases from 0.3381 to 0.3934, and the F1-score increases from 0.5054 to 0.5646.

Figures (8)

• Figure 1: Framework of the MRIFE. The two branches independently extract features and the decoder recovers the input images after feature fusion. A specially designed joint loss guides the updating of the two branches where the three encoders have the same network architecture but do not share parameters.
• Figure 2: (a) Original sample with $512\times512$ pixels; (b) part of original sample with $64\times64$ pixels; (c)-(e) masked block with $8\times8$ pixels: non-landslide block, landslide edge block, and landslide interior block.
• Figure 3: (a) Landslide similar to the background; (b) label; (c)-(e) side view/top vie/front view of the landslide in Google Earth.
• Figure 4: Visualized results of the comparative experiments. (a) Input; (b) label; (c)-(e) prediction: baseline, ICSSN, and MRIFE. The red line indicates the boundary.
• Figure 5: Grad-CAM results of the comparative experiments. (a) Input; (b) label; (c)-(e) heat map: baseline, ICSSN, and MRIFE. Red line indicates the boundary.