【摘要】 以柑橘为供试对象,研究了计算机视觉分级中的柑橘图像采集和图像信息处理方法。利用单个摄像机对每个柑橘从4个侧面采集4幅图像,利用4幅图像特征的平均值作为该样本的最终特征,研究基于果实颜色和表面纹理的柑橘图像快速处理分级技术。以人工感官鉴评结果作为分级标准,以训练好的概率神经网络对60个柑橘样本进行了分级试验。结果表明:系统对各级水果的识别成功率略有不同,对一级果和四级果的识别率最高,二级和三级果的识别率稍低,整体识别率约为85%。更多还原

【Abstract】 The technology of image grabbing and image processing for citrus grading was studied.Four pictures were grabbed for each orange from different perspective,a grading algorithm based on citrus color and texture features of four profiles was developed to identify the grade of the citrus efficiently.The experiment take the result of sense evaluation with manpower as grading criterion,and sixty patterns were graded with probabilistic neural network(PNN).The test result showed that the whole discriminating rate was up to 85%,and the rates of first class and fourth class were higher than the rates of second and third classes.更多还原