瓜果分类BP算法研究与实现  

The System Design of Image-based Fruit Classification

作　　者：李仕强[1] 傅德胜[2] 王水平[2] 赵英男[2] 

机构地区：[1]南京信息工程大学信息与控制学院,南京210044 [2]南京信息工程大学计算机与软件学院,南京210044

出　　处：《农机化研究》2010年第3期55-57,61,共4页Journal of Agricultural Mechanization Research

基　　金：国家自然科学基金项目(60702076);南京信息工程大学科研基金项目(Y636;07CX0016)

摘　　要：分析了图像的颜色空间模型,以HSI为基础模型,对采集到的瓜果图像进行去噪、分割、边缘提取等预处理。在此基础上,对水果图像的特征以及相关参数进行测量,最后采用BP神经网络及K-近邻规则对目标进行识别。实验证明,BP算法的识别率较高,系统总体识别率达到了98.92%,完全达到了应用需求。The color space model is analyzed in this paper. Based on HSI model, the pretreatment methods includes fruit image noise removing, segmentation, and edge extraction were done. Then the features and parameters were measured. At last we use BP neural network and K - Nearest Neighbor （KNN） rules to recognize the goal. The experimentation indicates that the recognition percent of BP neural network is very high. The recognition rate is about 98.92 percent, which reaches the application requirements.

关 键 词：瓜果分类 HSI BP神经网络 K-近邻 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]