指纹鉴定结论的概率化表述

The Research on the Probabilistically-represented Conclusion of Fingerprint Identification

目的基于指纹的5类纹型、9种细节特征点所占的概率及5类特征组合概率构建计算模型,用出现相同概率表述指纹鉴定同一结论。方法将选取指纹样本纹型分为5类、细节特征分为5类,根据5类特征中一、二、三类特征"连"与"非连"的属性划分9种细节特征点。结果统计分析全国三年比中2万余人份指纹样本5类纹型、细节特征点出现概率存在稳定的规律;选取广东、江苏、河南、青海、黑龙江五省指纹数据4 400余万人份,统计分析AFIS识别纹型的出现概率,佐证样本指纹5类纹型出现概率具有稳定规律。构建检材指纹出现概率计算模型,构建检材指纹5类特征在样本指纹5类特征中可能出现的最大值计算模型,验证概率化表述的指纹鉴定结论是否具有唯一性。结论基于构建的计算模型,计算出检材指纹与样本指纹相同的出现概率,实现概率化表述指纹鉴定结论的目的。

Objective Based on the probabilities of five types of fingerprint patterns, nine kinds of fingerprint minutiae and five sorts of feature combinations, a calculation model was to set up for representing the conclusion of fingerprint identification by the same probability between one test fingerprint and the sampled counterpart. Methods Five types of fingerprint patterns were defined along with the five sorts of features into which nine kinds of fingerprint minutiae were divided when the ＂connection＂ and ＂disconnection＂ were emerging among the first, second and third of the five-type features. Results Through more than twenty thousand persons＇ ten-digit fingerprints collected nationwide in three separate years, Lhe stable occurrence of probabilities was observed from the five-type fingerprint patterns and five-sort features. Over forty-four million persons＇ ten-digit fingerprints selected in Guangdong, Jiangsu, Henan, Qinghai and Heilongjiang provinces were carried out the statistical analysis on their probabilities of fingerprint patterns, demonstrating true of occurring the above-observed stability of probability. The calculation model was built for the probability of the test fingerprint whose five-sort features were thus able to show themselves with possible maximum in the sampled counterpart, thereby verifying the uniqueness ofprobabilistic representation of fingerprint identification conclusion. Conclusion With the same probability obtained by the established calculation model between the test fingerprints and sampling ones, fingerprint identification conclusion can be represented by the deduced probability.