Time-Efficient Object Recognition in Quantum Ghost Imaging

Abstract

Acquiring information at the fastest possible rate is often desirable,particularly in quantum ghost imaging which suffers from slow reconstructionspeeds. Many computationally intense deep-learning methods have beenimplemented in an effort to speed up image acquisition times by retrievingimage information. Often over-looked, machine learning methods can offerthe same, if not better, speed up in image acquisition time by an objectrecognition process. Four machine learning algorithms are implemented andtrained on a uniquely generated, noised, and blurred dataset of numericaldigits 1 through 9. Of the tested recognition algorithms, logistic regressionshows a 10× speed up in image acquisition time with a 99% predictionaccuracy. Additionally, this reduction in acquisition time is achieved withoutany image denoising or enhancement prior to recognition, thereby reducingtraining and implementation time, as well as the computational intensity ofthe approach. This method can be implemented in real-time, requiring only1/10 th of the measurements needed for a general solution, making it ideal forquantum imaging and recognition of light sensitive structure.