Automated Identification and Separation of Touching Rice Grains with Machine Vision

Automated identification and separation of touching kernels have important effects on result of quality detection of rice kernel with machine vision. This paper analyzed profile shape of isolated rice kernel and touching kernels and developed an algorithm based on the curve. The algorithm determined the curve and its direction which can reflect sharp degree of boundary. It can find quickly whether rice kernels touched or not and separated points. The method of minimum distance was used to segment touching kernels. For three types of rice kernels, algorithm could perform over 99% accuracy for identifying touching or not and can perform over 95% for segmenting.