Multi-scale Context Feature Based Change Detection Method for Remote Sensing Images

Abstract

Remote sensing image change detection plays an important role in many fields such as land detection, disaster assessment and urban construction planning. However, most of change detection methods just identify the change region but no the change type. In order to obtain the change region and the change type simultaneously, we propose a multi-scale context feature based change detection method in this paper. And the key part of the method is the multi-scale context aggregation network, which is composed of feature extraction module, attention mechanism module and spatial pyramid pooling module. The feature extraction module is based on HRNet, and we use partial convolution with depthwise separable convolution to reduce the parameters of the model and match the limited dataset. The attention mechanism module is used to efficiently select distinguishing features and the spatial pyramid pooling module is used to strengthen the ability of the model to segment objects with different scales. The experiments demonstrate the superiority of our method and the accuracy of the change detection is up to 86%.