煤堆图像分割与特征提取

Feature extraction based on image segmentation of coal pile images

基于煤堆图像分割存在重叠煤粒的边缘错综复杂、不易检测的问题,通过多种图像增强和边缘检测方法的效果对比,提出一种将对比度受限自适应直方图均衡法(CLAHE)和SUSAN边缘检测算法相结合的方法来检测煤堆图像中的煤粒边缘,并利用数学形态学和孔洞填充算法得到最佳种子区域,有效防止分水岭算法的过分割和欠分割现象,最后统计并分析煤粒分割区域10个特征参数的分布情况,包含了煤粒数量、大小、形状、颜色和纹理特征。研究结果表明:通过这些特征参数可以预测相关煤质信息,利于实现自动控制煤炭的分选。

Considering the complicated edges and difficult detection of overlap coal particle in the image segmentation of coal piles, several image enhancement and edge detection methods were contrasted. An algorithm combined with contrast limited local histogram equalization （CLAHE） and SUSAN edge detection was used to detect the edges of coal particles. Mathematical morphology and filling algorithm were used to obtain the best seed regions, which was able to prevent over-segmentation and under-segmentation effectively. Finally ten features’ distribution of coal segmentation regions, including features of coal number, size, shape, color and texture, were calculated and analyzed. The results show that the related coal quality information can be predicted using these features, which will be of great importance in automatic control of coal preparation.