A Machine Vision Approach for Grain Quality Control During Separation

The research aims to develop a method for determining the content of impurities in grain materials by analyzing digital images. The addressed problem was establishing a method for grain quality control using machine vision to create integrated automation systems for separating grain materials in post-harvest processing. Based on the research result, it was established that grain materials are characterized by significant heterogeneity in physical and mechanical properties and variability of their fractional composition due to changes in the conditions of their threshing by combined harvesters. The contamination of grain materials with non-grain impurities is 0.2–3.3 %. Therefore, to increase the efficiency of post-harvest processing, it is necessary to ensure fast and accurate control of grain material flows. The developed software performed segmentation and determined the pixel areas of grain material objects, providing the rejection of noise and false objects. The weight coefficients of the five main components of wheat grain materials were evaluated based on the experimental results. Statistical processing showed a relatively low variability of the weight coefficients of winter wheat grain, not exceeding 1.6 %. The developed method for evaluating the quantitative content of fractions in grain material allowed for determining the contamination of grain in the separation processes using the experimentally evaluated relative weight coefficients of the main five fractions for wheat grain materials. The developed method helps create control and automation systems for post-harvest processing of grain materials.