Research on Real-Time Pedestrian Detection Based on Infrared-Visible Image Fusion
Ziqin Shang, Baoping Cheng, Xiaoyan Xie, Tao Fu, Zijian Wu · 2023
With the advancement of AI technology and hardware, intelligent surveillance systems are becoming increasingly widespread, bringing more and more complex application scenes. Pedestrian detection, as one of the tasks within such systems, faces more challenges consequently. Current pedestrian detection methods primarily rely on visiblelight modality, limiting their performance and robustness. Introducing an additional modality, such as infrared, to improve performance imposes higher demands on platforms and resources. Therefore, multi-spectral image fusion emerges as a promising technique to enhance pedestrian detection. This paper proposes a pedestrian detection framework based on infrared-visible light image fusion. Additionally, three fusion methods are employed and compared against single-spectral results using existing detection models, which validate the potential for improving pedestrian detection performance.