基于视觉的草莓品质特征提取与评价模型

Vision-based strawberry feature extraction and quality evaluation model

为提高草莓分选效率、增加草莓品质评价方法的数据区分度,提出一种基于视觉的草莓品质评价模型。以行业标准的草莓品质传统评价方法为基础,通过图像处理方法提取草莓颜色、尺寸和形状,从而对草莓的成熟度、缺陷、质量和形状特征进行量化,实现对草莓的等级评价和规格评价。进一步采用熵权法,引入权重因子,建立草莓品质综合评价模型,实现对草莓品质的自动分级。设计草莓品质分选装置,验证该模型的有效性。测试结果表明,草莓等级和规格自动分级准确率分别达96.7%和98.4%,品质综合分级准确率达到96.3%,草莓分选成功率达到95.3%,每幅草莓图像品质分级平均耗时37 ms,每颗草莓分选平均耗时0.87 s。研究为基于机械手的草莓自动分选提供参考。

In order to improve the efficiency of strawberry sorting and increase the data differentiation of strawberry quality evaluation methods,a vision-based strawberry quality evaluation model was proposed.Based on the traditional evaluation method of strawberry quality of the industry standard,the color,size and shape of strawberries were extracted by image processing methods,so as to quantify the ripeness,defects,quality and shape features of strawberries,and realize the grade evaluation and specification evaluation of strawberries.Furthermore,the entropy weight method was adopted,the weight factor was introduced,and a comprehensive evaluation model of strawberry quality was established to realize the automatic grading of strawberry quality.A strawberry quality sorting device was designed to verify the effectiveness of the model.The test results show that the accuracy of automatic grading of strawberry grade and specification was 96.7%and 98.4%,respectively,the accuracy of comprehensive quality grading reaches 96.3%,the success rate of strawberry sorting reaches 95.3%,the average time of quality grading of each strawberry image is 37 ms,and the average time of each strawberry sorting is 0.87 s.This study provides a reference for manipulator-based automatic strawberry sorting.