HSONet:A Siamese foreground association-driven hard case sample optimization network for high-resolution remote sensing image change detection

TL;DR

HSONet tackles hard-case sample learning in high-resolution remote sensing change detection by integrating a Siamese foreground-scene association framework with an equilibrium optimization loss. The method jointly models scene context to reinforce background hard cases and dynamically balances foreground accuracy with background hard-case learning, using a variant FPN encoder and a lightweight dual-branch decoder. Empirical results on four public datasets show state-of-the-art performance, particularly in hard-case scenarios, supported by ablations validating the EO-loss and FS-relation module. The approach offers robust, scalable improvements for RS-CD and suggests potential extensions to semi-supervised settings.

Abstract

In the later training stages, further improvement of the models ability to determine changes relies on how well the change detection (CD) model learns hard cases; however, there are two additional challenges to learning hard case samples: (1) change labels are limited and tend to pointer only to foreground targets, yet hard case samples are prevalent in the background, which leads to optimizing the loss function focusing on the foreground targets and ignoring the background hard cases, which we call imbalance. (2) Complex situations, such as light shadows, target occlusion, and seasonal changes, induce hard case samples, and in the absence of both supervisory and scene information, it is difficult for the model to learn hard case samples directly to accurately obtain the feature representations of the change information, which we call missingness. We propose a Siamese foreground association-driven hard case sample optimization network (HSONet). To deal with this imbalance, we propose an equilibrium optimization loss function to regulate the optimization focus of the foreground and background, determine the hard case samples through the distribution of the loss values, and introduce dynamic weights in the loss term to gradually shift the optimization focus of the loss from the foreground to the background hard cases as the training progresses. To address this missingness, we understand hard case samples with the help of the scene context, propose the scene-foreground association module, use potential remote sensing spatial scene information to model the association between the target of interest in the foreground and the related context to obtain scene embedding, and apply this information to the feature reinforcement of hard cases. Experiments on four public datasets show that HSONet outperforms current state-of-the-art CD methods, particularly in detecting hard case samples.

Figures (17)

• Figure 1: Sample display of hard case sample in remote sensing image. Where hard case samples such as target occlusion, shadows and small targets are shown in red, yellow and orange boxes respectively.
• Figure 2: Overview of the HSONet.
• Figure 3: Schematic diagram of the FS-relation module.
• Figure 4: Scene information embedding schematic.
• Figure 5: Lightweight decoder architecture diagram