基于计算机视觉的橙子分级研究设计

Design of Oranges Based on Computer Vision

橙子具有很高的营养价值,是我国部分地区的支柱产业,在农业经济中占有一定地位。分级是水果销售前的一道重要工序,有利于增加销售收益,提高产业竞争力。橙子种类繁多,品质各异,对其检测分级显得尤为重要。目前,橙子的分级主要由人工完成,但不能满足可持续发展的要求。基于计算机视觉的橙子分级设备类型、分析和计算方法都较多,但处于试验阶段,还没有应用于实际生产。为此,基于计算机视觉技术,建立了橙子的实时分级系统。橙子图像用计算机进行预处理和灰度化后提取目标轮廓,然后分别对大小、颜色和表面缺陷进行检测,采用RBF神经网络模式划分等级。系统对各级橙子的识别准确率为82.5%~90.0%,平均准确率为8 6.3%。系统处理单张图像平均用时0.6 s,分级效率达到1 8 0个/s,可以实现对橙子的自动化检测和分级。

Oranges have a high nutritional value,is a pillar industry in some areas of our country,occupies a certain position in the agricultural economy.Grading is a fruit before the sale of an important process,is conducive to increasing sales revenue,improve industrial competitiveness.A wide variety of oranges,different quality,its classification is particularly important.At present,the grading of oranges is mainly done manually,but can not meet the requirements of sustainable development.Computerized visualization of orange grading equipment types,analysis and calculation methods are more,but in the experimental stage,has not been applied to the actual production.Based on the computer vision technology,this paper establishes the real-time grading system of oranges.The orange image is pretreated and grayed out by computer to extract the target contour,and then the size,color and surface defects are detected respectively,and the RBF neural network model is used to divide the level.The accuracy rate of the system is 82.5% ～ 90.0%,the average accuracy is 86.3%.System processing single image with an average of 0.6 s,grading efficiency of 180/s,can achieve the orange of the automated detection and classification.