摘要
为实现超市中水果蔬菜等产品的自动销售,提出了一种基于视觉特征的水果蔬菜自动分类方法。首先将所获得的水果蔬菜图像划分为多个重合的子块;接着提取这些子块的视觉特征,即尺度不变特征和方向梯度直方图特征;为了提高特征的表征能力,还将这些提取出的特征融合在一起描述目标;然后对融合后的特征做编码和池化操作以降低特征维数并提高特征区分能力;最后用所得特征训练支持向量机分类器并最终实现水果蔬菜的自动识别与分类。与现有方法相比,提出的方法在超市农产品数据库上取得了较高的识别率,从而为实现水果蔬菜的自动销售提供了技术支持和理论保障。
This paper presents an automatic method of fruit and vegetable classification,which is based on appearance features to make contributions to automatic selling of these goods in supermarkets.First,the fruit and vegetable images are divided into overlapping regions,followed by the appearance features extraction,such as Scale-invariant feature transform and Histogram of Oriented Gradient.Second,the features are fused together for improving the representative ability.To reduce the feature dimension and enhance the discriminative power,the coding and pooling processes are performed on the fused features.Last,a Support Vector Machine classifier is designed based on the features and then the fruit and vegetable can be automatically classified and recognized.Compared to the state-of-the-art works,the presented method produces a higher recognition rate and provides a possible solution for the automatic selling of fruits and vegetables in supermarkets.
出处
《重庆师范大学学报（自然科学版）》
CAS
CSCD
北大核心
2016年第3期115-120,共6页
Journal of Chongqing Normal University:Natural Science
基金
甘肃省高等学校基本科研业务费项目
甘肃省高等学校科研资助项目(No.2015A-087)
关键词
水果蔬菜图像
自动分类
视觉特征
编码
池化
fruit and vegetable images
automatic classification
appearance features
coding
pooling
