Extraction and modeling of carrot crack for crack removal with a 3D vision

文献类型： 外文期刊

第一作者： Xie, Weijun

作者： Xie, Weijun;Xie, Weijun;Yang, Deyong;Huang, Kai;Wei, Shuo

作者机构：

关键词： Carrot crack; Segmentation; Deep learning; NURBS; Multi -objective genetic algorithm

期刊名称：COMPUTERS AND ELECTRONICS IN AGRICULTURE （ 影响因子：7.7； 五年影响因子：8.4 ）

ISSN： 0168-1699

年卷期： 2024 年 224 卷

页码：

收录情况： SCI

摘要： The removal of defects from fruit and vegetable is conducive to improving resource utilization and reducing environmental pollution. However, it has been neglected by researchers. In this study, a novel method based on the deep learning (DL) algorithm and Non -Uniform Rational B -spline (NURBS) optimized by multi -objective genetic algorithm (GA) was proposed for extracting and modeling of carrot cracks. The crack segmentation model named Res -U -net was constructed with the pre -trained ResNet-50 and U -net. It performed best with pixel accuracy (PA), mean pixel accuracy (mPA), mean intersection over union (mIoU), and size of 98.48 %, 96.73 %, 92.17 %, and 0.24 G, respectively. Three-dimensional (3D) modeling of carrot crack provides a basis for feedrate scheduling and path interpolation in subsequent crack removal. For carrot crack modeling, the NURBS optimized by multi -objective GA was designed to fit the crack sliced point clouds for getting 3D smooth fitting curves with parsable expressions of carrot cracks in feedrate scheduling and path interpolation of the crack removal. The GAfull performed best with a reduction of approximately two-thirds in the data amount while maintaining the crack shape. The proposed extraction and modeling methods for carrot crack removal may offer a fresh thought into the postharvest treatment of fruit and vegetable.

分类号：

  • 相关文献
作者其他论文 更多>>

微信小程序