Study On Analysis Of Forensic Science Trace Paint Evidence With Pattern Recognition

This paper takes Fourier-transform infrared(FTIR) spectra of trace paint evidence which appear most frequently in all kinds of cases as the object of study, applying the method of pattern recognition SIMCA and the BP(Back Propagation) artificial neural network (ANN) algorithm which adapt to high dimension chemical quantity measuring data processing to trace paint evidence. Through the applying SIMCA method to the FTIR spectra of the trace paint sample for the clustering analysis after the clustering of SIMCA, the recognition rate and the rejection rate of the same kind of paint from different manufacturers are over 90%, obtaining the approving effect. Adopting three layers of BP network, ANN pattern is established to apply to the FTIR spectra pattern recognition of the paint sample. Matlab5. 2 artificial neural network toolbox is adopted to design three kinds of ANN recognition program. From the result of recognition, we know that the recognition capability of nonlinear-linear model of ANN, whose number of node in hidden layer is three, is the best. The correct rate of recognition can reach 97%. This paper makes a grouping study about the application of pattern recognition in the forensic scientific trace evidence analysis, and studies how to get hidden useful information by analyzing samples of trace evidences, so as to realize the classification trace evidence identification on the base of computer pattern recognition, and improve the accuracy and availability of trace evidence identification. A new method on analyzing difficult trace evidence has been established to solve the actual problem in the examination of trace evidence. The method of pattern recognition which uses FTIR spectra data to make analysis and identification for trace paint evidence is established. The result is accurate and feasible. It is the favorite new method for the identification of the forensic scientific trace evidence.The application of pattern recognition technology in the identification of the forensic scientific trace evidence not only are valuable in theory, but also have great practical application foreground. Trace evidence itself is a kind of complicated mixture, so it can only realize the differentiation of the class through using simple method to judge their all kinds of spectrograms. But the hairlike differences between the evidence of the same class are difficult to distinguish fast and exactly. So it is worthy and significant in practice and in terms of theory to introduce the method of data process of pattern recognition.