Abstract:
In order to sort the wheat kernels into different types automatically, a novel approach which integrates binary tree and support vector machine (SVM) is proposed to discriminate between four different types of wheat kernels by impact acoustic signals. At first, the impact acoustic signals were analyzed and potential features were exacted from them in both time and frequency domains. Then the SVM based on binary tree was used for pattern recognition. Detection accuracy rates of the presented system for undamaged kernel, insect damage, moldy and sprout damage were above 84.0%. The experimental results show that our research has a high value on application and provides a feasible method for automatic classification of wheat kernels.
