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

发布时间：2018-10-29 19:31

【摘要】：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.
[Abstract]:The accurate location of line in line tool trace image is the key to realize the automatic checking and comparing of criminal tool trace. Because the edge detection method is easily affected by image noise to produce false edge, it can not accurately locate the actual line in the trace, which leads to the difficulty of line feature extraction. In this paper, a new method of line location based on second derivative edge fusion combined with image context is proposed. First of all, according to the current line tool trace manual comparison method, the image inspection line is defined, the second derivative of the image is calculated along the alignment line direction, and the derivative value is fused regularly according to the property of the edge of the second derivative. To calculate the ridge and valley areas in line traces; Then according to the experience of manual line trace alignment, the line area is optimized, and the pseudo-line in the deep valley is deleted. Finally, the precise location of the line is realized by combining the image context to calculate the salient points in the ridge region. On the basis of the accurate location of the on-line strip, the experimental samples are registered with the second derivative absolute maximum ridge line, the grayscale maximum ridge line, the deepest valley feature line, and the distance between the feature lines as the feature template (vector). It can achieve a higher registration rate for the marks formed by the tool with obvious features of the working site. The experimental results show that the line location and registration method proposed in this paper can quickly realize the accurate location and registration of the lines in the line tool trace image, which lays a foundation for the automatic alignment of the line tool trace.
【作者单位】：广东警官学院计算机科学与技术系;广东警官学院刑事科学技术系;
【基金】：国家自然科学基金项目(61471134) 公安部应用创新项目(2015YYCXGDST026) 广东警官学院创新团队项目(2015JSTD03)
【分类号】：TP391.41

【相似文献】