基于改进分水岭-凹点分割的矿石粒径分级检测方法

Ore Particle Size Classification Detection Method Based on Improved Watershed-Concave Point Segmentation

摘要:

为了提高混凝土行业的生产质量，需要对矿石大小做粒径分析，传统方法是采用人工筛分处理，过程中需要耗费大量的人力物力，同时，也存在检测时间长和检测精度低等问题；针对这一难题，通过利用计算机视觉技术，提出了一种基于改进分水岭-凹点分割的矿石粒径分级检测新方法；首先，利用图像自适应中值滤波和改进的多尺度形态学处理，提取矿石轮廓特征；其次，采用改进的分水岭分割和凹点分割相结合，获得矿石之间粘连形成的深凹点集合；最后，引入反向链码模板对凹点集进行有效的分离，从而对矿石粒径做出精准的统计分析；实验结果表明，该算法的粒径分级与人工筛分的粒径分级相比较，两者之间的累积误差率在5%以内，具有较高的准确性与实用性，值得大力的推广与应用。

Abstract:

A particle size analysis of the ore size is required with a view to improving the production quality of the concrete industry. The traditional method is to use manual sieving processing, which requires a lot of labor and material resources. At the same time, there are also problems such as long detection time and low detection accuracy; To address this problem, a new approach to ore particle size classification detection based on improved watershed-concave segmentation is proposed by using computer vision technology. Initially, an adaptive median filter and improved multi-scale morphological processing are used to extract ore contour features. Secondly, the combination of improved watershed segmentation and concave point segmentation is used to obtain the set of deep concave points formed by adhesions between ores. Finally, an inverse chain code template is introduced to effectively separate the set of concave points to make an accurate statistical analysis of the ore grain size. According to the experimental results, the cumulative error rate between the particle size classification of this algorithm and the particle size classification of manual sieving is within 5%. Therefore, this algorithm has high accuracy and practicality, and is worthy of vigorous promotion and application.