U-Net-Based Optical Camera Communication in Intensity-Constrained Scenarios

Optical Camera Communication (OCC), which utilizes the unlicensed visible light spectrum, is regarded as a key enabler for future sixth-generation wireless technologies. In OCC, the communication performance, specifically the maximum achievable distance and bit-error rate (BER), relies on the pixel values captured by the image sensor. However, spatial and temporal intensity variations across pixels in rolling shutter-based OCC systems present a significant challenge for data decoding. This study investigates a low-illumination indoor environment where overlapping dark and bright intensity histograms are heavily impacted by noise, further complicating the data extraction process. To address this issue, this paper proposes UOCC, a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">U</i>-Net-based <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">OCC</i> architecture for image segmentation, designed to extract data under extremely low-light conditions in intensity-constrained scenarios by framing data pattern detection as pixel classification tasks. We further incorporate bit-plane slicing-based pre-processing and test-time augmentation-based post-processing to enhance the reliability of the model. Experimental results show that the proposed UOCC method achieves an excellent BER of <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">5 × 10⁻⁴</b> and a 97.20% intersection-over-union score after 50 epochs, significantly outperforming Otsu thresholding <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">(10⁻¹)</b> and both second- and third-order polynomial fitting methods <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">(3 × 10⁻²)</b> in this context. This performance is achieved by reducing noise through the exclusion of the lowest-order bit-plane of test images, which significantly enhances system reliability. Furthermore, the proposed method presents substantial potential for future applications, including non-line-of-sight and underwater communication environments.