基于改进YOLOv5的螺旋选矿机矿物分带图像分割算法研究

Research on Mineral Zoning Image Segmentation Algorithm for Spiral Concentrator Based on Improved YOLOv5

螺旋选矿机是一种流膜类重力选矿设备,目前其精矿的截取是通过工人观察矿物分带,根据经验确定精矿与中矿或尾矿的边界分割位置,并相应调节截取器的分矿块的分割点到精矿边界分割线位置,从而实现对精矿的准确截取和精矿品位的控制。由于每个工人的经验和技术水平不一样,难以保证每次获取的矿带分割线位置信息和调节操作的准确性,而容易造成选矿指标的波动。螺旋选矿机分选流体存在流速快、矿带边界模糊的问题,采用常规的图像识别算法和原始YOLOv5算法都难以得到满意的结果,针对此问题,提出了一种能够识别模糊小目标矿带分割点的改进YOLOv5算法,利用本算法对从工业中采集的螺旋选矿机生产矿带图像样本进行了矿带边界分割识别试验和测试。结果表明,改进的YOLOv5算法比原始YOLOv5算法准确度提高了14.3%,其识别的精度可以满足生产中对螺旋选矿机矿物分带自动识别的要求。

The spiral concentrator is a flow film type gravity concentrator.Currently,the interception of concentrate is achieved by workers observing the mineral zoning,determining the boundary division position between the concentrate and the middling or tailings based on experience,and adjusting the division point of the separator's ore block to the concentrate boundary division line position accordingly,thereby achieving accurate interception of concentrate and control of concentrate grade.Due to the different experience and technical level of each worker,it is difficult to ensure the accuracy of the location information and adjustment operation of the ore belt segmentation line obtained each time,which is easy to cause fluctuations in mineral processing indicators.The separation fluid of a spiral concentrator has problems of fast flow velocity and blurred ore belt boundaries,and it is difficult to obtain satisfactory results using conventional image recognition algorithms and the original YOLOv5 algorithm.In response to this problem,an improved YOLOv5 algorithm that can identify fuzzy small target ore belt segmentation points has been proposed.Using this algorithm,ore belt boundary segmentation recognition experiments and tests were conducted on ore belt image samples collected from industry using spiral concentrators.The results show that the accuracy of the improved YOLOv5 algorithm is 14.3%higher than that of the original YOLOv5 algorithm,and its recognition accuracy can meet the requirements for automatic recognition of mineral zoning in spiral concentrators in production.