A Change Detection Method Based on Multi-Scale Adaptive Convolution Kernel Network and Multimodal Conditional Random Field for Multi-Temporal Multispectral Images

Multispectral image change detection is an important application in the field of remote sensing. images usually contain many complex scenes, such as ground objects with diverse scales and proportions, so task expects feature extractor superior adaptive multi-scale learning. To address above-mentioned problems, a multispectral method based on kernel network multimodal conditional random (MSAK-Net-MCRF) proposed. The (MSAK-Net) extends encoding path U-Net, designs weight-sharing bilateral path, which simultaneously extracts independent features bi-temporal without introducing additional parameters. A selective convolution block (SCKB) that can adaptively assign weights designed embedded MSAK-Net to extract images. retains skip connections embeds upsampling module (UM) attention mechanism decoding give map better expression information both channel dimension spatial dimension. Finally, (MCRF) used smooth results MSAK-Net. Experimental two public datasets indicate effectiveness robustness proposed when compared other state-of-the-art methods.