Automated Bullet Striation Identification Technology Based On Image Processing

The increasing of gun crime cases bring a huge threaten to society stability,the policemen need to compare the evidence bullets with the bullets database tofind out the corresponding firearms and its owners in crime investigation. Todayin our country, such comparison is performed manually under a comparisonmicroscope. This approach can be time consuming, low efficiency andsubjectivity. It cannot meet the needs of rapidity and accuracy of crimeinvestigation. With the development of the computer assisted informationprocessing technology, it is possible to identify the bullets automatically. Thus,in this article a research on the technology of computer assisted bullet striationidentification is performed.In this article, the foundation and development status at home and abroadof the bullet identification is introduced. Then the two of the famous automatedidentification systems: DRUGFIRE and IBIS are also introduced. And afteranalyzing the characters of two typical algorithms of bullet identification:identification based on bullet striation distribution and identification based oncross-correlation function, the automated bullet striation identification systembased on image processing is established which includes image processingsection and cross-correlation operation section.According to the LEA data characteristics, the rough bullet striationinformation is extracted initially by making use of interpolation, Gaussian filterand mathematical morphology operation. And then the effective striation islocated by using edge detection, edge filter and Hough transform, at the sametime, the rifling angle is calculated, and the parameter of the LEA image qualityis presented. After getting the bullet effective striation by a mask, a compressedprofile is generated by averaging all valid data points vertically for each pixelcolumn. So the similarity of LEA compressed profile between evidence bulletsand reference bullets is calculated by applying cross-correlation function. Withtwo LEA compressed profile in the case of the same, similar, different, thecross-correlation values are calculated, and the results are very different.In the end, according to the designed identification system, softwareincluding bullet LEA image processing and cross-correlation operation isdeveloped under MATLAB development environment. With the software, anexperiment which tests the24LEA data of4bullets fired form2firearms isbuilt, the experiment results show that the cross-correlation values of the bullets fired from the same firearm are much larger than those not. And the resultsdemonstrate the effectiveness of the automated bullet striation identificationsystem based on image processing.