基于机器视觉的蔬菜种子包衣品质鉴定  

Quality Recognition for Coated Vegetable Seed Based on Computer Vision

作　　者：喻志成[1] 赵春宇[1] 高璐[1] 

机构地区：[1]上海交通大学电子信息与电气工程学院,上海200240

出　　处：《农机化研究》2017年第12期26-31,共6页Journal of Agricultural Mechanization Research

基　　金：国家"863计划"项目(2012AA10A505)

摘　　要：蔬菜种子包衣工作参数的智能调节,能提高包衣加工效率和成品质量。为了研究包衣工作参数的智能调节,提出了基于机器视觉的蔬菜种子包衣品质鉴定方法。针对蔬菜种子包衣过程中种子包衣完整性、包衣颜色深浅、包衣颜色均匀性3个重要指标,提出依据单粒种子的种子包裹率、种子颜色及纹理特征将包衣种子分为合格与非合格两类。对于种子图像中粘连的问题,采用分水岭算法将图像分割为单粒种子。通过对单粒种子的多阈值分割,实现种子包衣完整率的计算。基于HSI颜色空间提取H、S分量的颜色矩特征与I分量的灰度共生矩阵特征,融合种子包衣完整率、颜色矩特征和灰度共生矩阵特征这3种特征为一个11维特征向量,构建基于径向基核函数的支持向量机分类器对包衣结果进行品质鉴定。实验选用包衣后辣椒种子验证算法,结果表明:包衣结果识别准确率为90.93%。该研究可为后续研究包衣机工作参数的智能调节奠定理论基础。Intelligent regulation of vegetable seed coating parameters can improve the efficiency of the coating process, and improve the quality of the seed products.In order to study the intelligent control of coating parameters, this paper proposed vegetable seed quality identification method based on computer vision.According to three important indicators including coating rate, coating color and texture features, a seed was classified as qualified-coated or non-qualified-coated.For adhesion of the seed image, watershed algorithm was used to segment the seeds into single.Seed coating rate was calculated based on multi-threshold segmentation method.The seed image was converted into HSI color space in order to extract color moments features in H and S component and gray-level-co-occurrence matrix features in I component.Three kinds of features including coating rate, color moment features and gray-level-co-occurrence matrix features were fused into an eleven-dimensional feature vector.Support vector machine classifier was trained based on radial basis function.Experiments were conducted using coated chill seed.And results showed that the quality recognition for coated vegetable seed accuracy is 90.93%.This research laid a theoretical foundation of following-up study of intelligent regulation of vegetable seed coating parameters.

关 键 词：蔬菜 种子包衣 品质鉴定 机器视觉 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]