<p>Optical Camera Communication (OCC) has emerged as a viable alternative to radio frequency communication for line-of-sight applications. However, its commercialization remains limited, particularly in non-line-of-sight (NLoS) indoor environments. This paper presents the design of an OCC system capable of operating in NLoS conditions with potential industrial applications. The system employs an LED as a transmitter, modulating light using Camera On-Off Keying modulation. The modulated light reflects off rough surfaces and is captured by a camera acting as a receiver. Traditional NLoS OCC systems suffer from signal distortion, multipath interference, high path loss, and nonlinear distortions, leading to high error rates. To address these challenges, a deep convolutional neural network is utilized to detect and track reflected light sources with around 90% accuracy at a 1.2-meter distance under three different exposure times. Additionally, a UNet model reconstructs the stripe pattern from detected regions, achieving 95% accuracy through semantic segmentation-based preprocessing. The system demonstrates a bit error rate of 10⁻⁴ at 3.9 ms of exposure time in a Python environment. The proposed approach enhances the robustness and reliability of NLoS OCC systems, paving the way for commercialization in industrial and smart home applications.</p>