基于多特征融合与SVM的苹果品种快速识别算法

Fast Recognition Algorithm of Apple Varieties Based on Multi Feature Fusion and SVM

为快速无接触地识别苹果品种,采用机器视觉方式,提出基于多特征融合与SVM的苹果品种快速识别算法。首先,基于HSV和La*b*颜色空间提取仅含色度信息的H,S,a*,b*分量,通过直方图变换得到苹果颜色特征,同时考虑苹果表皮局部与全局纹理特性,研究基于GLCM与LBP的苹果纹理特征描述子,并通过串接方式进行颜色与纹理特征融合,实现不同品种苹果表面特征的准确描述;然后,构建基于SVM的苹果品种识别模型,并根据样本特征量优选线性核函数,实现苹果品种的快速准确识别;最后,通过实验验证,该算法兼顾精度和实时性,在测试集上平均准确率达94.0%,且推理时间仅2 ms,具有较高的实用价值。

In order to identify apple varieties quickly and contactless,a fast classification algorithm based on multi feature fusion and SVM is proposed by using machine vision.Firstly,based on HSV and La*b*color space,the H,S,a*and b*components with only chromaticity information are extracted,and the color features of apple are obtained by histogram transformation.Considering the local and global texture characteristics of apple skin,the apple texture feature descriptors based on GLCM and LBP are studied,and the color and texture features are fused by concatenation to realize the accurate description of the surface features of different apple varieties.Then,an apple variety recognition model based on SVM is constructed,and the linear kernel function is selected according to the features number of training samples to realize the fast and accurate apple variety recognition.The experimental results show that the algorithm takes into account both accuracy and real-time performance,with an average accuracy of 94.0%in the test set,and its recognition time is only 2 ms,which has a high practical value.