基于贝叶斯统计推断的粗集料级配测试研究

Research on coarse aggregate grading test based on Bayesian statistical inference

拌合站现场粗集料来料的检测一般是通过筛网筛分的方式进行,检测效率低下。因此,引入一种基于贝叶斯统计推断的机器视觉检测方法,实现对集料的快速筛分。利用工业相机采集下落过程中的集料颗粒并进行图像预处理,选取每颗集料的图像序列中最大的Feret短径作为图像特征。此特征为集料颗粒的局部特征,无法代表颗粒的全局三维特征。因此引入贝叶斯统计推断的方法,首先利用贝叶斯公式计算出每颗集料在不同实际粒径区间的后验概率,然后通过概率累加获得被测集料在不同集料规格的可能颗粒数量。为推断每颗集料级配占比,利用贝叶斯推断的方法计算后验分布,从而估计集料颗粒的占比情况。实验结果表明,预测粒径与实际粒径的分布占比最大偏差为1.67%,符合现场快速检测要求。

The detection of coarse aggregate incoming material in mixing station is generally carried out by the means of sifting of screening mesh,which has low detection efficiency.Therefore,the machine vision detection method based on Bayesian statistical inference is introduced to realize the rapid screening of aggregates.An industrial camera is used to collect aggregate particles in the falling process and perform image preprocessing.The largest Feret short diameter in the image sequence of each aggregate is selected as the image feature.This feature is the local feature of aggregate particles and cannot represent the global three‐dimensional feature of particles.Therefore,the Bayesian statistical inference method is introduced.The Bayesian formula is used to calculate the posterior probability of each aggregate in different actual particle size intervals.The possible particle quantity of the measured aggregate in different aggregate specifications is obtained by means of the probability accumulation.In order to infer the proportion of aggregate particles,the Bayesian inference method is introduced to calculate the posterior distribution,so as to estimate the proportion of aggregate particles.The experimental results show that the maximum deviation between the predicted particle size and the actual particle size distribution is 1.67%,which meets the requirements of on‐site rapid detection.