Volume estimation of unbroken soybeans samples using digital image processing techniques

Abstract

The calculation of volume of different oilseed grains, through computational models, has demonstrated effectiveness and efficiency. In the present work, the model has been extended to allow calculations of the volume of soybeans. In the model is proposed each grain of the sample is assimilated to a parallelepiped with main axes L (length), W (width) and T (thickness). The L and W values are determined from the Feret distances of the image, and the thickness is assumed to be proportional to width of the grain. The proportionality constant k is calculated using the formula of the model and validated against the experimental volume of the samples, fielding a confidence or percentual relative deviation. The model developed approximates soybean volume with confidence of a 1.25% using low-cost hardware for image acquisition and moderate computational resources.