果实形状的计算机识别方法研究

Studies on Technique of Computer Recognition of Fruit Shape

摘要: 在综合分析果实形状的基础上，提出用6个特征参数表示果形，这6个特征参数分别为半径指标、连续性指标、曲率指标、半径指标的对称性、连续性指标的对称性、曲率指标的对称性。首次将参考形状分析法用于果形判别，利用人工神经网络对果形进行识别和分级。结果表明，用提取的特征参数和果形识别技术，计算机视觉与人工分级的平均一致率在93%以上。

Abstract: Based on overall analysis of apple shape, originally,the reference shape method was used in apple shape feature extraction. The six shape feature parameters are Radius index, Continuity index, Curvature index, Symmetry of radius index, Symmetry of continuity index and Symmetry of curvature index. The method of using 6 shape feature parameters was used to describe the fruit shape. The 6 shape feature parameters were put into artificial neural network to train and recognize. The results showed that the method worked well with high accuracy in fruit shape sorting.