Recently, the Geometrical Computing and Intelligent Media Technology Team of the International School of Information Science and Engineering have made two latest scientific research achievements in the field of multi-source image fusion. Based on the practical problems faced by autonomous unmanned systems in complex and severe scenes, a new method has been proposed to solve a series of difficult problems of low-level image alignment and high-level semantic perception based on multi-source image fusion. The research results were published in CVPR, a top conference on computer vision, IJCAI, a top conference on artificial intelligence, and Information Fusion, a district I journal recommended by Chinese Academy of Sciences (CAS).
In complex and severe scenes, driven by the bottleneck of mono-modality sensor suffering from serious information loss, multi-modality imaging has gradually become a major technology in the field of autonomous driving. Infrared and visible imaging, which attracts the most attention, not only helps to improve the image quality in severe scenes, but also greatly promotes the downstream semantic perception tasks. The team's researches run through the whole path of imaging correction of multi-modality image, multi-modality information fusion and downstream semantic perception tasks, and puts forward innovative methods to solve the problems of multi-modality image misalignment, multi-modality image fusion focusing on visual quality and ignoring semantic transmission, and image fusion and multiple downstream semantic perception tasks not adapting to each other.
IJCAI (International Joint Conference on Artificial Intelligence) is one of the top international academic conferences in the field of artificial intelligence, and is also a CCF recommended Class A conference. The current IJCAI (2023) conference received 4,566 paper submissions, and the acceptance rate is only 15%.Information Fusionis one of the top international journals in the field of computer science, dedicated to disseminating the latest research and development trends in the field of information fusion, with an impact factor of 17.56, and is recommended by the CAS as a district I journal.
