Abstract: The lubricating oil pressure of ball mill in ore dressing plant is a critical real-time monitoring parameter. Bilinear interpolation method, DoG filter, Marr-Hildreth edge detector and watershed algorithm are studied on FPGA. In parti-cular, the algorithm HVPP is proposed. Segmentation, extraction and number identification are realized for the pressure pointer dial image by using above algorithms, such that intelligent reading for the gauge is achieved. Experimental result shows that employing the fast algorithm based on logic circuit to accomplish the machine recognition has advantages of good real-time performance and high recognition rate.

Key words: pointer dial, Field Programmable Gate Array（FPGA）, Marr-Hildreth edge detector, watershed algorithm, HVPP

摘要： 摘  要：选矿厂球磨机润滑油压力是一个至关重要的实时监测参数。基于FPGA采用了双线性插值法、DoG滤波器（高斯差分滤波器）、Marr-Hildreth边缘检测器及分水岭算法，并提出一种新算法——横竖端点寻点法（HVPP），对压力表图像进行分割、提取及数字识别，从而实现了对指针表盘图像的智能读数。实验结果表明采用基于逻辑电路的快速算法实现表盘图像的机器识别具有实时性好、识别率较高的优点。

关键词: 指针表, 现场可编程逻辑门阵列（FPGA）, Marr-Hildreth边缘检测器, 分水岭算法, 横竖端点寻点法（HVPP）