LAUSR

The complex vector beams yield up an abundance of polarization information that has not yet been well utilized in information encoding. In this paper, we propose a polarization encoding scheme with the non-orthogonal polarization states using a stationary vector beam. Recognizing those non-orthogonal polarization states is assisted by the structured patterns of the single vector beams under different polarization projections. We show that one can achieve different capacities of encoding bits by changing the step of the polarization angle with the single vector beam. We also demonstrate the non-orthogonal polarization encoding scheme can be well decoded with the machine learning classification algorithm. A 64×64 gray image is successfully transmitted by using 4 bits/symbol encoding-decoding scheme with 99.94 % transmission accuracy. Besides, by extending the encoding-decoding scheme to 8 bits/symbol based on the same single vector beam, we achieve a higher transmission rate with 65.58% transmission accuracy. Our work holds promise for small-angle non-orthogonal polarization encoding for free-space optical communications.