Machine Vision‐Based Morphological Benchmarking and Extrusion Optimization of Analog Rice: A Digital Image Analysis Framework for Agricultural Grain Quality Evaluation

ABSTRACT The appearance of analog rice strongly influences product acceptance. This study developed a low‐cost Python machine‐vision workflow to characterize, benchmark, and optimize the morphology of extruded analog rice. A total of 420 grains were analyzed using 10 dimensional, shape, and curvature descriptors. Measurement validation used 40 paired images, comprising 20 Basmati and 20 optimized analog rice grains, measured in ImageJ and Python. Analog rice was produced using a custom‐made single‐screw extruder under a 6 × 6 factorial arrangement of screw speed (25–50 Hz) and cutter‐speed (45–70 Hz) settings. Each combination was manufactured once, with 10 grains treated as within‐run subsamples. A fixed two‐factor model and quadratic response‐surface models were used to evaluate the observed operating window. Screw speed promoted grain elongation, whereas cutter speed had a stronger influence on width, aspect ratio, CI, BA, and total morphology error. All CI values satisfied CI ≥ 1. Analog‐rice main‐effect means were 1.0055–1.0066 for CI and 1.80°–2.90° for BA. Within the tested range, 50 Hz screw speed and 45 Hz cutter speed produced the best dimensional match. The optimized grains achieved 96.6% similarity to Basmati based on eight‐dimensional and shape descriptors. CI and BA were evaluated separately as straightness indicators. The moderate length and total‐error model fits, boundary solution, and unreplicated manufacturing runs limit extrapolation. The workflow provides an objective and affordable method for at‐line grain morphology evaluation.