基于改进神经网络算法的蔬菜图像识别

Vegetable image recognition based on improved neural network algorithm

文章提出了改进神经网络算法,建立了径向基函数神经网络模型,包括梯度下降方法求解权重参数,增大邻域半径的均值聚类方法求取隐函数中心值,利用相邻聚类中心获得核宽度,通过量子遗传算法删除冗余权重和神经元;提取了蔬菜图像的特征,并给出了算法流程。仿真试验表明,试验算法对蔬菜图像的形状特征平均识别率为97.56%,纹理特征平均识别率为95.60%,颜色特征平均识别率为93.25%,训练时间平均为5.83s、识别时间平均为2.18s,优于其他算法。

In order to improve the effect of vegetable image recognition,a neural network algorithm was proposed.Firstly,radial basis function neural network model was established,including gradient descent method for solving the weight parameters.Kmeans clustering method increasing the radius of neighborhood was calculated implicit function center value,and the nuclear width was used adjacent cluster centers.Secondly,quantum genetic algorithm was deleted the redundant weights and neuron.Thirdly,vegetable image feature extraction was extracted.Finally,the process was given.The simulation results showed that the average recognition rate of shape feature was 97.56%,and the texture feature was 95.60%.Moreover,the color feature was found 93.25%,after trained for 5.83 s,and the average recognition time was 2.18 s.The algorithms we reported here was found better than other kinds.