likelihood coefficient

Introduction. The current task of forensic ballistic identification is to investigate the breech face impressions on the firing cartridge cases. The paper deals with comparison of images of breech face impressions by correlation cells method. Theoretical analysis. Similarity of images of investigated impressions can be established by a method of the correlation analysis. However, this method is ineffective, because the correlation coefficient is dominated by areas that do not contain useful information. To improve the efficiency of correlation analysis, J. Song proposed a method of congruent matching cells. The method of correlation cells is based on this method. The paper presents the algorithm of the method, broken down into the main stages. The first stage includes preliminary processing of the image and splitting it into cells. The second stage is building the distribution of the maximums of correlation coefficients. The third stage is to establish the presence of the cluster in the graph of this distribution. The fourth stage allows to identify areas valid for further work. The method of selecting areas that do not contain breech face impressions is based on finding the autocorrelation function. In the fifth stage, the final value of the correlation coefficient is calculated without taking into account the areas of the analyzed images that do not contain useful information. Experimental research. Two conditions were proposed as criteria for similarity of the breech face impressions. The first one is a high value of the correlation coefficient. The second is the presence of the cluster on the graph of maximum correlation coefficient distribution. The verification of this fact is based on the identification of the likelihood coefficient. Conclusion. The method of correlation cells made it possible to increase the efficiency of the correlation analysis. In the course of the work, the criteria for attributing impressions to the category of similar ones were defined.