“京蜜八号”哈密瓜成熟等级分类研究—基于机器视觉和神经网络  

The Mature Grading of‘JingMi No.8' Hami Melon—Based on Computer Vision and Neural Network

作　　者：胡光辉[1] 郭俊先[1] 虞飞宇[1] 李俊伟[1] 刘军[1] 刘亚[1] 

机构地区：[1]新疆农业大学机械交通学院,乌鲁木齐853002

出　　处：《农机化研究》2015年第2期190-194,共5页Journal of Agricultural Mechanization Research

基　　金：新疆维吾尔自治区教育厅基金项目(XJEDU2013I14);国家自然科学基金项目(61367001)

摘　　要：目前,新疆哈密瓜在采摘期成熟等级判定主要采用人工方式,容易混淆级别,影响销售质量。为此,提出了一种基于机器视觉和主成分分析优化神经网络的哈密瓜成熟等级识别方法。首先,利用田间哈密瓜图像采集系统获取了采摘前不同成熟期的哈密瓜图像,利用图像处理技术获得了感兴趣区域;其次,提取能表征哈密瓜不同成熟等级的外观特征,包括H分量图像的色调累积频度、纹理特征、几何特征;最后,利用主成分分析法优化特征,构建并验证基于BP神经网络的哈密瓜成熟等级预测模型。研究表明,基于机器视觉和主成分分析优化神经网络预测哈密瓜成熟等级是可行的,准确率达86.59%。At present, the ripeness judging of the picking Hami melon is to rely mainly on manpower in the Xinjiang Ag- ricultural, it is easy confusing about the grade, and affecting the sales quality. For identification the mature grade of Ha- mi melon quickly, the method based on the computer vision and the optimized Principal components neural network was proposed. First, obtained the Hami melon images of the different ripeness using of an acquisition system based on com- puter vision , and extracted the region of interest using of the digital image processing technology. Second, extracted the external characterization information that can be reflective of the different mature grade of the Hami melon, including the cumulative frequency of Hue,the textural feature and the geometric features. Finally, optimize feature based on PCA , built and verified the neural network model, predicted the different mature grade of Hami melon. Research shows that identification and detection of the mature grade of Hami melon based on machine vision and the optimized neural network using PCA is feasible, the recognition rate was 86.59%.

关 键 词：哈密瓜 成熟等级 机器视觉 BP神经网络 主成分分析 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]