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.