星光级宽光谱全彩色成像技术研究是针对当前夜间由于环境光照度低, 传统的可见光成像、激光成像、热成像等技术方法只能获取目标的灰度信息, 而不能获取目标的真彩色信息。据统计, 彩色图像包含的信息是灰度图像的30倍, 光谱信息在目标识别中极为关键。文中采用一种集超大像元、超薄工艺微透镜集光、大口径通光的成像方式, 并在此基础上设计深度学习图像处理算法, 对器件物理真彩色进行彩色增亮增强, 进一步提升图像信背比。最终将器件的感光能力提升50~100倍, 彩色夜视最低照度0.000 4 Lux@25 Hz, F1.0, 从而实现宽光谱全彩色成像, 为后续目标检测、识别和产业化应用提供技术基础。

Research on starlight level wide spectrum full color imaging technology is aimed at the current demand of obtaining the real color image in the low ambient light for other technical methods such as traditional visible light imaging, laser imaging and thermal imaging, which can only obtain the target grayscale imaging information not the target true color information. According to statistics, color images contain 30 times information than that of grayscale images, and spectral information is extremely critical in target recognition. An imaging method with super large pixels, ultra-thin process micro lens collection and large-diameter light transmission is adopted. Based on this, a deep learning image processing algorithm is designed to enlighten and enhance the physics true color of the devices, and the signal-to-background ratio of images is increased further. Lastly, the sensitivity of the device is increased by 50~100 times and the minimum illumination of color night vision is 0.000 4 Lux@25 Hz, F1.0, thereby wide spectrum full color imaging is realized, which provides technical bases for target inspection, recognition and application.