基于笔痕角度特征对高水平机器人书写字迹的识别研究

Research on Recognition of High-level Cartesian Robotic Writing Based on Stroke Angle

本文研究了高水平写字机器人书写字迹特征,建立了一种根据笔痕角度识别高水平写字机器人书写字迹的方法,提出在检验中区分此类字迹与人自然书写笔迹的对策和注意事项。通过使用奥博8000笔迹量化检验仪的层析和超景深技术对高水平写字机器人书写的样本和人自然书写的样本进行测量,获取笔画墨迹宽度和深度,计算笔痕最深处与笔画边缘的水平距离和笔画墨迹宽度的比值i,再运用图表和差值运算分析i值分布的离散程度,分析高水平写字机器人样本与人自然书写样本的笔痕角度的聚集性和波动性。结果表明,高水平写字机器人书写的样本笔痕角度值聚集度高且波动性小,而人书写的样本笔痕角度值聚集性低且波动性大,因此,笔痕角度特征是识别高水平写字机器人书写字迹的有效方法。实验中还发现,长笔画的书写角度更具有稳定性,转折笔画和交叉笔画的书写角度数据会受到干扰,因此选取长笔画进行测量比较准确度更高。

Whether the suspicious handwriting is written or not is the key in the handwritirg examination,which directly determines the conclusion.The traditional methods of forging handwriting,inclicding printing,copying and stamping,are easy to identify,because there are no stroke pressure.But the advances in robotic writing technology pose new challenges for forensic document examiners.Robotic writing can form three-dimensional stroke indentation,and it is easy to misjudge the handwriting formed by normal writing in the examination.Some researchers have modified the Cartesian writing robot to make its handwriting more natural,and more similar to the handwriting written by human,which is difficult to distinguish.Forensic document examiners should be alert to this kind of high-level forged handwriting.In this study,by utilizing the optical tomography and depth-of-field technology of AB 8000 handwriting quantitative tester,we measured the strokes of natural handwriting samples and high-level robotic writing samples,and obtained the width and depth of stroke,calculated the ratio i of the horizontal distance between the deepest mark and the edge of stroke and the width of stroke ink.We alsa measured the distribution degree of i value by chart and differential value operation,and analyzed the aggregation and fl uctuation of the mark angles between high-level robotic writing samples and human handwriting samples.The results showed that the stroke angle of high-level robotic writing samples has high aggregation and low volatility,while the stroke angle of human writing samples has low aggregation and high volatility.Therefore,it is verified the validity of stroke angle feature in recognizing the robotic writing.It is also observed in the experiment that the stroke angle of long strokes is more stable,and the angle data of turning strokes and crossing strokes will be disturbed,so it is more accurate to choose long strokes for measurement.