A research team led by senior engineer Li Jianping from the Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences has recently designed a deep learning image coloring algorithm, which can automatically color grayscale images of plankton taken in situ underwater very close to human observation.
The research findings were presented at the European Conference on Machine Vision, one of the three top international machine vision academic conferences Sunday, Shenzhen Evening News reported yesterday.
More plankton imaging systems have made color possible with its technological development. Meanwhile, many experiments have proven that color images can help observe more about planktons.
However, underwater color imaging requires white light, which leads to zooplankton crowding in front of the underwater imaging device due to phototaxis, changing their spatial distribution underwater. This kind of unnatural change will yield deviation from observation results on planktons.
According to Li, most plankton is not sensitive to long-wavelength red light, and most traditional underwater imaging systems use red light or near-infrared light to avoid zooplankton crowding due to phototaxis.
“However, such shooting conditions can only obtain grayscale images of plankton. It is an ingenious solution to transforming grayscale images obtained under red light illumination into high-fidelity color images by artificial intelligence (AI),” Li was quoted as saying.