一种指针式油位计读数识别算法

A Reading Recognition Algorithm of Pointer Type Oil-level Meter

针对现实生活中油舱油位由于油位计读数不准而出现油位亏空或者溢出的情况,提出了一种新的油位计读数自动识别算法。首先,在预处理阶段使用高斯型同态滤波器增强油位计图像对光照的适应性,并通过中值滤波、二值化以及图像形态学等操作突出油位计的指针区域;然后,根据油位计轴心颜色不同于黑色指针颜色这一特殊属性,遍历指针像素找出颜色发生突变的坐标点,进而准确求出油位计的轴心坐标;再根据轴心位置选定包含油位计指针的感兴趣区域,对感兴趣区域进行霍夫变换即可准确地检测出指针位置;最后,以红色为标准分割出油位计的起点和终点位置,进而运用角度法自动识别出油位计的读数。实验结果表明,该方法不但可以提高油位计自动读数的准确率,而且大大降低了识别油位计读数所需要的时间。

In the real life,due to the inaccurate readings of the oil-level meter,the oil level in the tank may be overfilling or in deficit.For this,we propose an automatic recognition algorithm for oil-level meter reading. Firstly the algorithm uses the Gaussian homomorphicfilters in preprocessing to improve the illumination adaptability of oil-level meter images,and then highlights the pointer area of oil levelmeter by the median filter,binarization and image morphology. Secondly,according to the special property that the color of oil-level me-ter axis is different from that of black pointer,the coordinates of the color mutation can be found by traversing the pointer pixels so as toprecisely determine the axis coordinate of the oil-level meter. Next,the ROI （region of interest） which contains the pointer of oil-levelmeter is selected according to the axis position,then the pointer position can be accurately detected by the Hough transform in the ROI.Finally,the starting and terminal position of the oil-level meter are divided by red as the standard,and then the angle method is used toautomatically recognize readings of the oil-level meter. The experiment shows that the proposed algorithm not only improves the auto-matic reading accuracy rate of the oil-level meter,but also greatly reduces the recognition time.