Research On Multi-spectral Technology For Criminal Investigation Evidence Analysis

Multi-spectral technology is a fast and non-destructive analysis and detection technology.Its multi-spectral image contains both the spatial position information of the measured object,as well as the radiation and absorption information of the measured object.Based on the application status and requirements of multi-spectral technology in the field of criminal investigation and evidence analysis,this paper is devoted to the application and technical problems in the field of physical analysis such as blood age estimation,blood trace detection and fingerprint segmentation by constructing a multi-spectral low-cost imaging system with LED as the active lighting source.In this paper,a visible-near-infrared multi-spectral imaging system with 8 LED as illumination source and black-and-white CCD camera as imaging unit is constructed.A multi-spectral software based on Windows platform and C# language is designed,which includes two functional modules of light source control and image acquisition.The module realizes the function of automatically switching the illumination source and capturing the needed image.The feasibility of accurately estimating the age of human blood using the multi-spectral system is studied,and the results are compared with those obtained by using hyperspectral technology.The prediction results of different algorithm models are studied.K-nearest neighbor algorithm,support vector machine algorithm,random forest algorithm and the fusion algorithm of these three models are used to model and compare.It is found that the fusion model has the best results.It also has better adaptability and accuracy for blood samples from different sources.The average error in 0-2 days is 0.057 days,and the correct classification rate is Corr.Ect classification rate(CCR)reached 75.6%,the average error was 0.467 days in 3-20 days,and the CCR reached 65.6%.The results are similar to those obtained by Hyperspectral techniques for predicting the age of blood stains.A method for detecting bloodstains on complex backgrounds using the above multispectral system is proposed.Identifying blood and non-blood on white paper,identifying blood on colored substrates and identifying blood on dark stained clothes are studied.Using the blood characteristic absorption band combined with the effective band in background recognition as the selection band increased the recognition rate of bloodstains on colored backgrounds.The use of grayscale expansion instead of increasing exposure time reduces the noise after brightening the image.Using false color synthesis instead of the traditional band fusion as an image enhancement and using SVM method improves the accuracy of the results.Using the boolean logic performs classification on the blood,instead of the traditional image enhancement to display the bloodstain,resulting blood trace recognition map is more discriminating.The experiment result shows that this method is accurate and the blood trace image is recognitive.In the experiment of identifying bloodstain on dark red clothing,the recognition accuracy rate reaches 97%.The method of fingerprint image segmentation using the multi-spectral system is studied.The fingerprint image segmentation method is proposed,which is based on the initial segmentation of the fingerprint image,the shape school,the secondary segmentation based on the support vector machine algorithm.By comparing and verifying the foreground and background blocks of the multi-spectral image of the artificial fingerprint samples as the evaluation criteria of segmentation error rate,it is found that compared with the conventional methods,the proposed fingerprint image segmentation method has a lower error rate,especially for the fingerprint samples with complex background,which is only 3.32%.