The Road to On-board Change Detection: A Lightweight Patch-Level Change Detection Network via Exploring the Potential of Pruning and Pooling

TL;DR

This paper introduces LPCDNet, a lightweight patch-level change detection network that efficiently filters unchanged patches in large satellite images, significantly speeding up processing and reducing memory usage on edge devices.

Contribution

The paper proposes a novel sensitivity-guided channel pruning and multi-layer feature compression approach to create a lightweight, high-speed change detection network for satellite imagery.

Findings

LPCDNet achieves over 1000 frames/sec on NVIDIA Jetson AGX Orin.

Reduces over 60% memory costs for pixel-level change detection.

Maintains comparable change detection performance to existing methods.

Abstract

Existing satellite remote sensing change detection (CD) methods often crop original large-scale bi-temporal image pairs into small patch pairs and then use pixel-level CD methods to fairly process all the patch pairs. However, due to the sparsity of change in large-scale satellite remote sensing images, existing pixel-level CD methods suffer from a waste of computational cost and memory resources on lots of unchanged areas, which reduces the processing efficiency of on-board platform with extremely limited computation and memory resources. To address this issue, we propose a lightweight patch-level CD network (LPCDNet) to rapidly remove lots of unchanged patch pairs in large-scale bi-temporal image pairs. This is helpful to accelerate the subsequent pixel-level CD processing stage and reduce its memory costs. In our LPCDNet, a sensitivity-guided channel pruning method is proposed to remove unimportant channels and construct the lightweight backbone network on basis of ResNet18 network. Then, the multi-layer feature compression (MLFC) module is designed to compress and fuse the multi-level feature information of bi-temporal image patch. The output of MLFC module is fed into the fully-connected decision network to generate the predicted binary label. Finally, a weighted cross-entropy loss is utilized in the training process of network to tackle the change/unchange class imbalance problem. Experiments on two CD datasets demonstrate that our LPCDNet achieves more than 1000 frames per second on an edge computation platform, i.e., NVIDIA Jetson AGX Orin, which is more than 3 times that of the existing methods without noticeable CD performance loss. In addition, our method reduces more than 60% memory costs of the subsequent pixel-level CD processing stage.