| In Chinese forensic science,it is an important part to analyze the three-dimensional footprints using the theory of three-dimensional footprints' pattern imprints.However in real cases,the usage of three-dimensional footprints' pattern imprints analyze is mainly rely on the experience of technicians.The disadvantage of subjective judgment still exists.In order to solve these problems and make the theory of three-dimensional footprints analysis more useful,this research took the relative methods in pattern recognition as main means,developed the detection algorithms of dig-imprints and tread-imprints in three-dimensional footprints and achieved the goal of automatic extraction and classification of three-dimensional footprints' pattern imprints.Firstly,a large number of photographs were taken using the way of criminal scene footprints extraction method.Then,cut out the dig-imprints and tread-imprints in the images.A data set of dig-imprints and tread-imprints was set up.Then,Haar-like and LBP were chosen as the methods of pattern extraction.The pattern vectors were extracted.The classification algorithm was Adaboost.And two detection algorithms based on Haar-like and LBP patterns were developed separately.The test TPR of these two algorithms were 98.00% and 100.00%.At last,the conclusion of was drawn by protracting the ROC curve and calculating the AUC value.After that,with the same way,the pattern vectors were extracted and the detection algorithm was developed.The TPR of the two algorithms were 91.00% and 93.00%.The conclusion that the algorithm based on Haar is better was drawn by protracting the ROC curve and calculating the AUC value.Through the research,the possibility of automatically extracting and classifying the pattern imprints of three-dimensional footprints was proven.And good results were achieved in the detection of dig-imprints and tread-imprints.This has provided a new horizon of identification recognition with pattern imprints in three-dimensional footprints. |
PDF Full Text Request