Vision-based real-time monitoring and signal stabilization for automated lotus fiber extraction process

ABSTRACT Lotus silk is a rare and sustainable textile produced from lotus stems; however, its manual extraction process results in low productivity and inconsistent fiber quality. Real-time monitoring of fiber extraction is essential for automation but remains unaddressed. This study proposes a real-time framework integrating image processing and hybrid data filtering to estimate equivalent fiber area during the extraction process. The system applies thresholding and Canny edge detection to quantify fiber area and length, followed by a hybrid Moving Average–Double Exponential Moving Average (MA–DEMA) filter to stabilize the noisy signals. Parametric analysis identified most appropriate values of the smoothing factor (α = 0.005) and window size (N = 101) to balance smoothness and responsiveness. Compared with MA, EMA, DEMA, and 1DKalman filters, the proposed hybrid model improved noise reduction by 41.6%, achieving Sm = 2.81, VRR = 0.51, and MAE = 408.3. The developed approach enables real-time monitoring and can be extended to automatic control of the lotus fiber extraction process, contributing to digitalization in natural fiber production.