摘要
【目的】研究涟红温州蜜柑pH的机器视觉检测及影响检测精度的因素。【方法】对机器视觉系统采集的柑橘图像进行图像裁切、RGB空间至HSI空间的转换和差值法去图像背景,用色调H和饱和度S为输入,建立小波神经网络柑橘pH预测模型,无损检测柑橘pH。【结果】30个测试样本的检测结果表明,预测偏差最大值为9.95%、偏差最小值为-3.6%、平均偏差为0.8%、标准偏差为2.95%,pH±0.1精度内的正确识别率为80%,pH±0.2精度内的正确识别率为93.33%。【结论】涟红温州蜜柑pH与果皮色泽之间具有相关性,可用机器视觉检测其pH。但进一步提高预测精度,首先须在图像处理环节上去除各种虫斑与病斑的影响。
[Objective] pH value measuring method of ＇Lianhong＇ citrus fruits based on machine vision and factors influencing measurement accuracy were studied. [ Method ] Images of citrus fruits from machine vision system were processed by cutting, converting from RGB space to HSI space, removing background by deviation. A wavelet neural network model was constructed to detect pH value of citrus fruits non-destructively, the inputs of the model were image hue H and saturation S. [ Result ] Results of test of 30 samples showed that the maximal deviation of pH value was 9.95%, the minimal was -3.6%, average deviation was 0.8%, standard deviation was 2.95%. The correctness of detection for accuracy ±0.1 and ±0.2 were 80% and 93.33%, respectively. [ Conclusion ] pH value has correlation to pericarp color and luster, and can be detected with machine vision method. To improve the detection accuracy, influence of speckles of insects and diseases should be removed during image processing.
出处
《中国农业科学》
CAS
CSCD
北大核心
2008年第11期3741-3745,共5页
Scientia Agricultura Sinica
基金
湖南省教育厅科学研究项目(06D059)
关键词
柑橘
PH
小波神经网络
图像处理
Citrus fruit
pH
Wavelet neural network
Image processing
