Aquila: A Hierarchically Aligned Vision-Language Model for Enhanced Remote Sensing Image Comprehension

Recently, large vision-language models (VLMs) have made marked strides in vision-language capabilities through visual instruction tuning, showing great promise in the field of remote sensing image interpretation. However, existing remote sensing vision-language models (RSVLMs) often fall short in capturing the complex characteristics of remote sensing scenes. While some recent RSVLMs, such as EarthGPT and LHRSBot, have started to incorporate multiscale features, they typically adopt relatively shallow fusion strategies and lack tightly integrated mechanisms for fine-grained vision-language alignment. In this paper, we present Aquila, an advanced vision-language foundation model designed to enable richer visual feature representation and more precise vision-language feature alignment for remote sensing images. Our approach introduces a learnable hierarchical spatial feature integration (SFI) module that supports high-resolution image inputs and aggregates multi-scale visual features, allowing for the detailed representation of complex visual information. Additionally, the SFI module is repeatedly integrated into the layers of the large language model (LLM) to achieve deep vision-language feature alignment, without compromising the model’s performance in natural language processing tasks. These innovations significantly improve the model’s ability to learn from image–text data. We validate the effectiveness of Aquila through extensive quantitative experiments and qualitative analyses, demonstrating its superior performance.