结合图像上下文的二阶导数边缘融合线条精确定位与配准

Precise Location and Registration of Striation by Integrating Image Context and Second Derivative Edges Fusion

线条工具痕迹图像中线条的精确定位是实现犯罪工具痕迹自动检验比对的关键.针对边缘检测方法易受到图像噪声的影响产生虚假边缘,不能够对痕迹中的实际线条进行准确定位,导致线条特征提取困难的问题,提出一种结合图像上下文的二阶导数边缘融合线条精确定位方法.首先根据当前线条工具痕迹人工比对方法定义图像检验的比对线,沿比对线方向计算图像的二阶导数,并根据二阶导数边缘的性质对导数值进行规则融合,以计算线条痕迹中的脊线和谷线区域;然后根据人工线条痕迹比对经验对线条区域进行优化规整,删除深谷中的伪线条;最后结合图像上下文计算脊线区域中显著点实现线条的精确定位.在线条精确定位的基础上,以二阶导数绝对值最大脊线条、灰度值最大脊线条、最深谷为特征线条,以及特征线条之间的距离为配准特征模板(向量)对实验样本进行配准,能够对工作部位具有较显著特征工具形成的痕迹实现较高配准率.实验结果表明,文中提出的线条定位和配准方法能够快速实现线条工具痕迹图像中线条的精确定位和配准,为实现线条工具痕迹自动比对奠定基础.

Precise location of striation in the image is a key technology to realize automatic examination andcomparison for criminal striation tool mark.At present,the edge detection methods are often applied to extractlines from tool mark.However,the edge detection methods are sensitive to image noise,and easily producefalse edges.The realistic striations are difficult to be located precisely and extracted by using edge detectionmethods for tool mark.The precise location approach of striation by integrating image context and second derivativeedge fusion was proposed in the article.Firstly,the comparison line was defined based on the methodof artificial identification.The second derivative of image was computed along comparison line.The ridge andgroove regions of striation mark were calculated by utilizing rules fusion for derivative values according to theproperties of second derivative edge.Then,the ridge and groove regions were optimized and normalized,falsestriations in the deep grooves were deleted as well,based on the experience of artificial identification.Lastly,the significant points among ridge regions were computed by integrating image context and the striations werelocated precisely.On the basis of striation locating precisely,the registration of experimental samples was performed using registration vectors such as ridges with second derivative maximal absolute values,ridges withmaximal gray scale,deepest grooves,as well as distances among them.The experimental results show that theapproach presented in the paper can locate the striations and make image registration precisely and fast.Higherrate of registration can be reached especially for striation marks produced by tools with prominent individualfeatures on the work side.Obviously,the proposed method is helpful to establish automatic comparison systemfor striation tool mark.