摘要
非固结颗粒的粒度分布在河床或海床稳定性分析、沉积物输移以及沿海扩散等方面发挥着重要作用.随着图像处理技术的发展涌现出了许多数字图像筛分方法.本文依据数据处理结果的维度将数字图像筛分方法分为二维数字图像粒径筛分方法和三维数字图像粒径筛分方法.首先,根据二维数字图像粒径筛分方法所依赖的核心算法将其进行了分类并总结了每种方法的适用性和不足.其次,分析了三维数字图像粒径筛分方法并总结了其存在的不足.最后,依据各个方法优缺点提出了"几何"粒度筛分方法和"人工智能"粒度筛分方法相融合的新构想,即前者为后者提供标记训练集,经过训练后后者可以快速适应新环境的数字筛分任务.
Abstract
The grain size distribution of unconsolidated particles plays a significant role in riverbed or seabed stability,sediment transport,and coastal dispersion.With the development of image processing techniques,many digital image particle size sieving methods have emerged.This paper classifies digital image screening methods into two-dimensional digital image particle size screening methods and three-dimensional digital image particle size screening methods based on the dimensions of data processing results.Firstly,according to the core algorithms of 2D digital image particle size sieving methods,they were classified,and the applicability and shortcomings of each method were summarized.Secondly,the three-dimensional digital image particle size screening methods were analyzed,and their shortcomings are summarized.Finally,based on the advantages and disadvantages of each method,a new concept of integrating"geometric"and"artificial intelligence"particle size screening method was proposed,i.e.,the former provides the latter with a labelled training set,and the latter can be quickly adapted to new environments for digital screening tasks after training.The task of digital sieving in new environments can be quickly adapted after training.
基金项目
山东省优秀中青年科学家奖励基金(BS2014SF016)
NETL
NSTL
万方数据