基于多示例学习的超市农产品图像识别

摘要

为解决超市农产品价格需依靠人工记忆的问题,实现农产品的智能识别,提出了基于多示例学习的农产品图像识别方法.采用改进的单色块及其邻域算法(SBN)特征提取算法将训练样本组织成多示例包,利用多样性密度算法对正包和反包进行多示例学习,根据多样性密度最大化模型对测试样本进行识别.分别在自采集的多类别果蔬图像集以及Amsterdam图像库中的单类别果蔬图像上进行测试.结果表明该方法能够识别不同光照、存在干扰物的环境背景下,以任意方式摆放的多类别混合果蔬图像,识别率最高达到94.21％,且对于单类别果蔬图像的识别优于全局方法.因此利用基于多示例学习的图像识别方法对超市农产品的自动售卖提供辅助具有可行性.%The pricing of agricultural products in supermarket needs to rely on artificial memory. In order to realize intelligent recognition of agricultural products, an image recognition method of agricultural products based on the multi-instance learning was proposed. An improved Single Blob with Neighbors (SBN) method was adopted to organize bags and meanwhile extract features of an image. The target concept was learned by maximizing Diverse Density (DD) and applied to images recognition. Experiments were performed on both multi-class produce image dataset by self-collection and single-class agricultural product images selected from Amsterdam Library of Object Images (ALOI). The experimental results show that the method is able to recognize multi-class agricultural product images captured under various illumination conditions and interference environment, and the recognition rate can achieve 94.21 percent. Additionally, the method performs better than global method when recognizing single-class agricultural product images.