基于图像特征的多支非电子类温度计刻度并行识别方法

Parallel recognition method for scales of multiple non-electronic thermometers based on image features

为了提高医护人员的工作效率,减少医护人员与患者的接触频率,降低其感染风险。以机器视觉为基础,提出了一种基于图像特征的非电子类温度计刻度识别方法。首先,根据像素点对采集到的图像进行预处理,为了对目标区域进行进一步处理,利用霍夫直线变换对图像进行旋转矫正;然后,通过分割算法提取图像中水银液柱和刻线,并由刻度线位置信息,通过像素点计算出体温计示数;最终实现了3支体温计并行的读数识别和记录,在35.5~40.5℃内相对误差不超过±0.35%。

In order to improve the work efficiency of medical staff and reduce the frequency of their contact with patients,reduce the risk of infection,based on machine vision,a scale recognition method of non-electronic thermometer based on image features is proposed.Firstly,perform preprocessing on the collected image according to the pixel points.In order to further process the target area,the image is rotated and corrected by the Hough line transformation.And then,the mercury column and scribed line in the image is extracted by segmentation algorithm and the thermometer readings are calculated by pixel point from the position information of the scale line.Finally,the reading recognition and recording of three thermometers in parallel can be realized,and the relative error does not exceed±0.35%in the temperature range of 35.5~40.5℃.