基于混合高斯图像窄带模型钨矿初选技术研究

针对钨矿石初选环节中人工手选作业效率低、成本消耗大等问题,提出机器视觉与图像处理技术相结合实时钨矿初选方法。通过引入GPU加速混合高斯模型进行矿石运动目标检测,提取图像前景中完整的矿石目标。结合图像信息,提出融合灰度特征与图像窄带的矿石目标识别算法,快速获取钨矿石中脉石的位置信息,为钨矿石的初步分选提供依据。试验结果表明,相比较传统人工手选作业方式,该方法极大提升矿石分选速度与精度,满足工业化实时检测识别要求。

Research on Primary Separation of Tungsten Ore Based on Mixed Gauss Image Narrow Band Model

Under the view of the low efficiency and high cost of manual hand separation in the primary separation of tungsten ore, a method of combining the machine vision and image processing technology to the real-time tungsten ore primary separation is proposed. Introduced the Gaussian mixture model (GMM) into moving target detection, the ore target in the foreground of the image can be extracted. Combining with the gray character information of the image, an algorithm of ore target recognition combining gray feature and narrow band is proposed. The position information of gangue in the tungsten ore is quickly obtained, which provides the basis for the preliminary separation of tungsten ore. The experimental results show that compared with the traditional manual hand separation mode, the method can greatly improve the speed and accuracy of ore separation, and meet the requirements of industrial real-time detection and identification.