基于LBP的粒子群优化混合核函数ELM的蔬菜水果图像分类方法

Research on Vegetable and Fruit Image Classification Method Based on LBP Particle Swarm Optimization Mixed Kernel Function ELM

蔬菜水果图像的分类研究在智慧农业领域具有重要的地位。针对蔬菜水果的实时高精度的分类问题,提出了一种基于LBP的粒子群优化混合核函数ELM模型(LBP-PSO-KELM)的分类方法,首先通过LBP等价变换提取蔬菜水果图像的纹理特征,然后将多项式核函数和高斯核函数加权成的复杂核函数(KELM)引入极限学习机中,并使用粒子群算法(PSO)对KELM中的核系数进行自适应选取,同时对引入的其他3个关键参数寻找最优值,获得最优模型。通过对Fruits-360数据集进行分类实验,结果表明,LBP-PSO-KELM在此数据集的分类准确率达到98.865 8%,平均分类时间为3.7ms,比单核ELM以及传统的分类方法分类准确率更高,模型计算时间短,且对硬件设备要求低,满足了智慧农业的实际需求。

The classification and recognition of vegetable and fruit images play a significant role in the domain of smart agriculture. To work out the issue of real-time high-accuracy classification of vegetables and fruits, this paper proposes a particle swarm optimization based on LBP mixed kernel function ELM model(LBP-PSO-KELM) classification method, first by LBP equivalent transform to extract the texture characteristics of fruits and vegetables;and then the polynomial kernel function and gaussian kernel function weighted as complex nuclear functions(KELM) are then introduced into the limit learning machine;the kernel coefficients in the KELM are adaptively selected using the particle swarm algorithm(PSO);at the same time find the optimal value for the other three introduced key parameters and obtain the optimal model.Through the classification experiment of Fruits-360 data set, the results show that the classification accuracy of LBP-PSO-KELM in this data set reaches 98.3658%. Compared with mononuclear ELM and traditional identification methods, the classification precision is higher. The model requires only a short computation period and does not require high speed computer hardware, which satisfies the practical needs of smart agriculture.