AI helps light source recognition, machine learning greatly reduces the number of measurements

- May 18, 2020-

The identification of light sources plays an important role in the development of many photonic technologies, such as lidar, remote sensing and microscopes. Traditionally, identifying diverse light sources such as sunlight, laser radiation, or molecular fluorescence requires millions of measurements, especially in low-light environments, which limits the implementation of quantum photonic technology. In the Applied Physics Reviews, researchers showed a smart quantum technology that can significantly reduce the number of measurements required to identify light sources.


The author of the paper, Omar Magana-Loaiza, said: "We trained an artificial neuron using the statistical fluctuation characteristics of coherent light and thermal light." After the researchers trained the artificial neurons with light sources, the neurons could identify potential features associated with specific types of light.


"One neuron is enough to significantly reduce the number of measurements required to identify the light source from millions to less than one hundred," said Chenglong You, the paper's corresponding author and researcher.


As the number of measurements decreases, researchers can identify the light source more quickly. In some applications, such as under a microscope, they can limit the damage of light, because they do not need to illuminate the sample nearly multiple times during measurement.


"For example, if you are doing an imaging experiment with delicate fluorescent molecular complexes, you can reduce the time the sample is exposed to light and minimize any light damage," another co-author Roberto de J . León-Montiel said.


Cryptography is another application that can prove the value of these findings. Typically, to generate a key to encrypt e-mail or information, researchers need to make millions of measurements. "We can use similar neurons to accelerate the generation of quantum keys for encryption," Magana-Loaiza said.


Because laser light plays an important role in the field of remote sensing, this work can also develop a new family of intelligent lidar systems that can identify intercepted or modified information reflected from distant objects. Lidar is a remote sensing method that measures the distance to the target by illuminating the target with a laser and measuring the reflected light with a sensor.


"With our technology, the probability of an intelligent quantum lidar system being disturbed will be greatly reduced." He said. In addition, the possibility of resolving lidar photons from ambient light such as sunlight will be of great significance for remote sensing at low light levels.