传统图像分割算法在农作物籽粒考种应用中的研究进展

Research progress of traditional image segmentation algorithm in seed testing of crops

传统图像分割算法以时间、空间复杂度低等优点在农作物籽粒考种领域中有着广泛的应用。对传统分割算法在农作物表型获取过程中的应用进行研究,首先阐述Otsu、分水岭、边缘检测、SLIC算法以及凹点分析算法的算法原理,对种皮颜色灰度均匀、形状不同的农作物籽粒,以“问题—方法”的模式阐述不同算法在应用中存在的问题以及相应的解决方法;接着将算法基于阈值、区域、边缘、聚类、凹点整合为五大类,对算法的分割效果、优缺点及其适用范围进行比较研究;最后,剖析农作物籽粒图像分割应用研究存在农作物种类覆盖度不够宽泛、图像分割精度不高、技术通用性不高等问题,并从算法精度提高、重叠遮挡处理等方面对未来的研究进行展望,以期为农作物籽粒考种过程中的图像分割研究提供参考。

Traditional image segmentation algorithm has been widely used in the field of crop seed testing because of its low complexity in time and space.The application of traditional segmentation algorithm in the crop phenotype extraction was studied in this paper.Firstly,the algorithm principles of Otsu,watershed,edge detection,SLIC and concave point analysis algorithm were expounded.For crop seeds with uniform seed coat color and different shapes,the problems in the application of different algorithms and the corresponding solutions were described in the model of‘problem-method’.Then the algorithms were integrated into five categories based on threshold,region,edge,cluster and concave point,and the segmentation effect,advantages and disadvantages and application range of the algorithm were compared.Finally,the problems in the application of crop seed image segmentation were analyzed,and the future research directions were prospected from algorithm accuracy improvement and overlapping occlusion processing,in order to provide reference for the research of image segmentation in the process of crop seed testing.