Design of Reconnaissance Image Compression System Based on Semantic Segmentation
Siyan Zhang, Hongxiang Chen, Zepeng Wang · 2021 IEEE International Conference on Unmanned Systems (ICUS) · 2021
In order to ensure the real-time processing of reconnaissance images and achieve complete and accurate acquisition of “high-value information” in images based on limited wireless channels, the paper proposes a scheme for reconnaissance image compression system based on semantic segmentation. The scheme proposes an image compression method: according to the task requirements, the method selects the corresponding semantic category of the target to perform semantic segmentation on the reconnaissance image, and performs hierarchical and lossless compression on the semantic segmentation result according to the importance of the target and then transmits it. And the method is combined with the existing direct compression and ROI-based image compression methods to achieve the purpose of obtaining high-quality reconnaissance information based on the limited wireless channel capacity under different mission requirements.