Instrumental Colour and Size Grading of Pulse Grains

Scientists at the Canadian Grain Commission have developed a machine vision system for grading pulse grains. This system utilizes a flatbed scanner and an image analysis based software developed at the Canadian Grain Commission. Features extracted from an image of a bulk sample are used for colour and size assessments. A neural network is developed to make a colour assessment while morphological image processing techniques are used for seed size measurements from the image. The machine-assessed colour matches the human inspectors more than 95%. The size measurements from the bulk image are also highly accurate (within 0.2 mm of the actual size).