非均匀光照下马氏瓶刻度图像的自动识别研究

摘要

马氏瓶刻度的自动识别会直接影响到潜水面蒸发量的观测与计算效率[1]。针对室内物理实验光照不均匀情况下的马氏瓶刻度图像的识别，首先使用改进的 Bernsen 算法对图像进行二值化处理；然后求得二值化图像的垂直投影曲线，利用 Loess （locally weighted scatterplot smoothing）算法对垂直投影曲线进行平滑处理，求得平滑处理后的各个极值点，计算出相邻极值点像素值之间的差值，最大差值对应的位置即马氏瓶液位线；根据得到的液位线和识别的数字计算出液位值，即可计算出潜水面的蒸发量。实验结果表明，改进的 Bernsen 算法改善了室内物理实验中非均匀照射下的马氏瓶刻度的分割效果。将此方法与可调焦的摄像头相结合，可实现潜水面蒸发量观测的自动化、连续化和智能化。此方法对类似的刻度识别具有重要的参考价值。%The automatic identification of the Marriotte’s bottle scale has direct influence on the observation and the computational efficiency of water table evaporation.To deal with the problem of identifying Marriotte’s bottle scale image in the condition of indoor experiments’uneven illumination,the first step is to use the improved Bernsen algorithm to let image binarized to get the vertical projection curve of the binary image and then smooth it by using the Locally Weighted Scatterplot Smoothing (Loess)algorithm to calculate the difference between the pixel value of the adjacent extreme points which were got after smooth processing,and the position of the maximal difference is the liquid level line of the Marriotte’s bottle.According to the liquid level line and the recognized numeral,the liquid level value and the water table evaporation can be calculated.Experimental results show that the improved Bernsen algorithm improves the result of segmentation of the Marriotte’s bottle scale in the condition of indoor experiments’uneven illumination.What’s more,with the combination of this method and manual-focus or auto-focus camera,it is possible to implement the automation,serialization and intelligence of the observation of water table evaporation.This method is an important reference to similar scale recognition.