Study Of The Firearm Identification For The Three-dimensional Surface Of Primer Based On Spline Filter And The Congruent Matching Cells

The toolmarks found on fired bullets and cartridge cases are the important clues and physical evidence for the criminal cases and judgments. The manufacturing processes, the erosion and abrasion during use will produce some unique features on the parts of firearms. The breech face impressions and firing pin impressions on the primer of cartridge cases are the major marks for firearm identifications. Thus, studying the automatic identification technologies based on the three-dimensional(3D) topography of primer is significant for improving the investigation.This dissertation studies on the formation and the acquisition method of the ballistic toolmarks and tests two groups of toolmark samples, then proposes two problems in primer identifications: The first is that there are no effective areal filter for extracting the ballistic features; the second is that there are no high accurate identification method for 3D surface.The existing areal filter can not process the low frequency curve due to the endeffects. Thus, for primer surfaces, especially for the sunken surfaces of firing pin impressions, a new areal filter should be built for extracting the ballistic features. In this thesis, a set of new boundary conditions are proposed for reducing the end-effects of the profile spline filter recommend by ISO16610-22(2006). By applying the new boundary conditions to the profile cascade approximating spline filter, this thesis builds an isotropic areal spline filter which can be widely applicable to complicated 3D surfaces.Because the valid data of breech face impressions and firing pin impressions are the non-rectangle data, large void area will attend the correlation. This thesis proves that the results of the classical Normalized Cross Correlation(NCC) will be affected by the void area, and proposes a new NCC function based on valid data(VDNCC) to eliminate the void area. Comparing other schemes, VDNCC can improve the accuracy of correlations and adapt the fast algorithm.The Congruent Matching Cells(CMC) theory proposed by NIST, suggests that the valid correlation area and the invalid correlation area should be isolated by cutting the sample data into correlation cells, and detecting the congruent between correlation cells to improving the capability of ballistics identification. The CMC method based this theory is built in this thesis. In the breech face identification tests, this method can identify all 780 test sample pairs. Compared with the NCC tests, the identification capability is improved.After filtering by the areal spline filter, the ballistic features can be extracted for correlation. Because the size of valid area size is smaller than breech faces and the similarity is also lower, the CMC method can get satisfactory results in firing pin identifications. An Improved CMC(ICMC) method is proposed by building “CMC-θ” distribution. Based on the test of 4005 pair of firing pin identifications and the comparison with the studies of FIB and NIST, the ICMC achieves more accuracy.In conclusion, this thesis implements the automatic identification technology and obtain satisfactory results by building areal spline filter and CMC method. The research results could improve the studies of 3D surface filtrations and stochastic topography matching.