Research On The Feature Extraction And Matching Technology Of Criminal Investigation Traces

With the development of the economy and the progress of technology,Social criminal cases have changed.Criminal means have become more intelligent,which has increased difficulties to solve the case.In the process of criminal investigation,the scientific extraction and full utilization of the trace texture left over from the crime scene help the judicial official to solve the case justly,quickly and scientifically.The traditional research on comparison of criminal traces is mainly done manually,which is time-consuming,inefficient and too subjective.The trace texture which can not be obtained by human eyes is extracted and analyzed by digital image processing technology,in which,the objective comparison and quantitative analysis of physical evidence trace can be realized.And it also provides a scientific basis for case detection.In this paper,in order to promote criminal trace extracting and comparing levels,the process ablity of X-10 B optical comparison microscope is upgrade by digital image processing technology.In order to improve the imaging effects of microscope,SIFT feature point matching technique is used to stitch image,in which,the observation field of microscope is expanded more than 2 times.Then,we propose a new method of generating whole area clear picture with large depth of field by merging some picture with different focus field: dividing normal picture into serval blocks,compare Laplacian energy of each picture in every block,select the highest one as the output block.As for linear trace,one of the most common trace in criminal investigation,a new criminal information extract method was designed for linear trace: using rectangular window to intercept the segment along the vertical direction of the trace line,then align and compare trace lines automatically.Another kind of common trace in criminal investigation is object contour,Through the Hu moment matching,we can achieve contour classification,in addition,a contour expansion algorithm is proposed to generate the contour curve to achieve a fine matching between highly similar contoursExperiments was also designed and done in this parper.Multiple sets of criminal trace samples were test in the experiments,the results of trace collection and trace matching are in good agreement with the expected results,and the effect is good.