基于L-M神经网络的苹果形状快速分级

The apple shape grading at high speed based on levenberg-marquardt neural network

为了实现苹果形状的有效分级,提出了一种快速的智能分级方法,即将苹果边缘曲线的傅立叶描述子作为果实形状特征,采用先进的Levenberg-Marquardt算法训练神经网络结构,最后用训练好的神经网络进行苹果形状分级。实际应用表明该方法切实可行且效果显著,不仅分级速度快,而且分级正确率高达90%以上。

In order to grade the apple with shape,a novel fast intelligent grading method is presented,which is based on the advaneod Levenberg-Marquardt algorithm to train the neural network. The main process is to take the Fourier description of the apple＇s edge curve as the shape features,then train the neural network architectures by L-M algorithm,finally grade the apple shape with the trained network. The actual application shows that the method can achieves high precision, and get very fast grading speed. In apple shape grading,the application effect is very notable.