Self-supervised image change detection method based on lightweight capsule network

In response to the significant impact of speckle noise on the detection accuracy of synthetic aperture radar (SAR) image changes, the high network model complexity of existing capsule network-based image change detection methods, and the loss of a large amount of original image information in training samples, this paper proposed a self-supervised image change detection method based on the light capsule network (SLCapsNet). The logarithmic ratio operator difference graph was generated, and the “pseudo label” of training samples with high confidence was obtained through the maximum inter-class variance method and fuzzy C-means clustering method, which laid the foundation for self-supervised learning. The paper constructed a three-channel training sample based on the two temporal SAR images and difference graph of logarithmic ratio operators to maximize the preservation of sample information. Lightweight capsule network was designed to extract training sample features through single scale convolution, and a single scale capsule network was used to mine spatial relationships between features. Comparative experiments and ablation experiments were set up, and tests were conducted on five real SAR datasets. The experimental results show that the advantage of the proposed method is to improve the operational efficiency of the method while reducing model complexity, obtain stronger robust features, suppress the adverse impact of speckle noise on change detection performance, and improve change detection performance.