罗志翔　胡蓉　赵全友　邓雄昌

摘要：近些年，计算机视觉发展迅速，在水果识别方向进行了广泛的应用和研究。本文设计基于BP神经网络的水果识别系统，选取生活中常见的三种水果：苹果、橘子、香蕉作为对象。首先，通过网络资源等搜集水果图像建立样本库;然后通过MATLAB对图像进行预处理，为后续的特征提取做好准备。水果特征的提取选择纹理、形状、颜色三种特征进行提取;同时在每种特征中选用不同的特征值作为特征向量。通过提取三种特征后输入到BP神经网络中进行训练、识别。经测试，识别的成功率可以达到93.18%，证明了可行性以及未来的可实用性。

关键词： 水果识别;MATLAB;特征提取;BP神经网络

中图分类号： TP391        文献标识码：A

文章编号：1009-3044（2021）04-0187-03

Abstract： Computer vision has developed rapidly in recent years， and has been widely used and researched in the direction of fruit recognition. This paper designs a fruit recognition system based on BP neural network. Choose three common fruits in life： apples， oranges， and bananas as objects. First， collect fruit images through network resources to establish a sample library. Then the image is preprocessed by MATLAB software to prepare for subsequent feature extraction. For the extraction of fruit features， three features of texture， shape， and color are selected for extraction; at the same time， different feature values are selected as feature vectors in each feature. After three kinds of features are extracted， they are input into BP neural network for training and recognition. After the recognition test， the recognition success rate can reach 93.18%， which proves the feasibility and practicality in the future.

Key words： fruit recognition; MATLAB; feature extraction; BP neural network

我國地域广阔、水果资源丰富，随着科技发展和社会需求的加大，水果的种植规模也在不断增大;同时也出现了复合型果园。在复合型果园场景中，借助水果识别系统配合机械化设备可以完成自动采摘、分拣等操作;尽管人工同样能实现相同的操作，但是效率不高、需要的劳动力也较大。而且，随着劳动力结构的转变，从事农业的劳动力也在不断减少。所以将其应用到生产当中可以代替大量的劳动力，可以达到高效、方便的特点[1]。

在当前，国内外对水果识别以及神经算法的研究已经取得了不错的成绩;在实际应用中也是不断地得到完善和提高。基于社会的发展和科技水平的不断提升，人民的需求也在不断变化，从而让物体识别技术也在不断发生改变。……