汕头大学学报（自然科学版）

对于文献中图的曲线数据的提取,提出一种通过采集像素点来识别曲线坐标的方法,并通过Matlab软件实现.首先用滤波器去除图片噪声,并确定用于灰度图转化为二值黑白图的阈值,然后通过搜索坐标框在灰度矩阵中的位置来确定真实坐标与灰度矩阵坐标的比例因子,最后获取曲线各像素点在灰度矩阵的坐标,并乘以比例因子得到曲线各点的坐标值.对曲线图像进行处理的结果表明,这种方法提取的数据准确,相对误差在-0.7%～+1.2%范围内,精度取决于原图像像素点的数目.

A method of data extracting through Matlab programming is introduced.First,the image was denoised by filter,and optimal threshold value was used to converted gray scale image to binary image.Then,scaling factor was obtained by determining the location of frame of axis in gray scale matrix.All features except curves were set to white color,so that the data of curve were easily recognized and converted to real coordinate.The method is used to extract two types of curve of picture and the results show that data obtained via this method are precise.The relative error is between-0.7% to 1.2% and the precision mainly depends on the number of pixel point.