摘要
建立准确高效的品质分级方法,对于花卉产业的标准化发展具有重要的意义.针对人工分级劳动强度大、经验要求高、缺乏统一性等问题,提出了基于机器视觉技术的盆栽微型月季品质分级方法.该方法根据盆栽微型月季的株高、花朵数、整齐度、花盖度和病虫害状况特征的不同,分别提取其相应的特征参数,并利用最小二乘支持向量机作为分类器,对其品质等级进行评价.经试验研究,单独利用株高、花朵数、整齐度、花盖度和病虫害状况特征进行分级的准确率分别为95.65%、94.68%、94.68%、94.20%、96.61%,而综合特征分级准确率为99.50%,验证了特征提取和分类模型的有效性,利用综合特征进行分级时间为10 s,明显提高了分级效率,该方法为建立准确高效的智能盆栽微型月季分级方法提供了理论参考.
Abstract
Potted mini rose has the characteristics of varieties,difficult of storage and large quality gap.Therefore,the establishment of accu-rate and efficient quality grading technology is of great significance for the standardized development of flower industry.As the traditional manu-al grading method need high labor intensity,high experience and lack of unity,we proposed a new quality grading method of potted mini rose based on machine vision.According to the different characteristics of plant height,flower number,uniformity,flower coverage and pest status of potted mini roses,the corresponding characteristic parameters were extracted,and the least squares support vector machine was used as a classi-fier to evaluate the quality grade.Through the experiment,the accuracy of a single characteristic were 95.65%,94.68%,94.68%,94.20% and 96.61%,respectively,while the accuracy of grading using the synthetic features was 99.50%,which verified the effectiveness of the proposed method.Generally,the manual grading method took several days,while the time of grading by the proposed method was 10 s,which obviously improved the efficiency.This method could provide basis for establishing an accurate and efficient intelligent grading method of potted mini rose.
基金项目
宁夏回族自治区自然科学基金(2022AAC03756)
宁夏回族自治区青年科技人才托举工程项目(第五批)()
维普
万方数据