基于图像分析的粗粒煤堆密度组成估计

Estimation of density distribution of coarse coal pile by image analysis

针对煤样密度组成人工测量的滞后性,提出了一种基于图像分析的粗粒煤堆密度组成实时估计方法.引入煤堆图像定向分割算法和全局分割算法,提取了50个煤粒表面特征参数,根据其随密度级的变化趋势筛选出了3个有效特征参数,利用改进的KNN算法预测煤粒密度级,并结合煤粒质量模型实时估计煤堆密度组成.测试结果表明,粗粒煤堆密度组成估计的绝对误差最高为7.15%,最低为1.41%.

An estimation of density distribution of coarse coal pile by image analysis was pro- posed because of the time-consuming of traditional density distribution measuring. Two image segmentation algorithms, directional segmentation and global segmentation, were introduced in this paper. Fifty surface features of coal particles were extracted and finally three ones were se- lected according to the change trends of features with the density change. The improved KNN algorithm was used to predict the coal density fractions, and then density distribution in real- time by virtue of mass model of coal particles was estimated. Tests indicate the highest abso- lute error of density distribution estimation is 7.15 %, and the lowest absolute error is 1.41%.