| Hyperspectral imaging technology is a technique that combines image information with spectral information.It can record spectral information in different wavelength ranges and perform qualitative and quantitative analysis of the inherent physicochemical properties and spatial distribution of samples.With its dual information of spatial geometry and spectral information,hyperspectral imaging technology has been widely used in fields such as agricultural product testing,geology,biomedical science,and forensic investigation.The identification and detection of bloodstain patterns,as well as the prediction of bloodstain exposure time,have always been important means for investigators to understand violent crimes.By effectively detecting and processing bloodstains at the crime scene,a large amount of physical evidence related to the criminal case can be obtained.Traditional bloodstain identification detection methods include phenolphthalein reagent method,luminol method,and ultraviolet detection method,etc.All of these methods have major limitations in crime scene application,and such destructive detection methods can make subsequent detection tasks difficult.With its fast and nondestructive detection capability,hyperspectral imaging technology effectively addresses the long detection time,low detection accuracy,and the phenomenon of missed detection in complex scenes that occur in traditional bloodstain detection tasks.This study aims to address the challenges of bloodstain detection in complex scenes,especially the challenges of bloodstain target detection and analysis.To this end,this paper proposes a bloodstain morphology feature recognition method and a bare exposure time detection method based on near-infrared highspectral imaging technology.The main research contents are as follows:1.A method of detecting morphological characteristics of bloodstains using visible and near-infrared hyperspectral imaging is proposed.In the course of this research,a set of hyperspectral imaging detection system is established to detect and identify bloodstains and bloodstain analogues using near-infrared hyperspectral imaging technology,and threedimensional data of the samples are successfully acquired.The task of identifying and detecting bloodstains in complex scenes and complex states is carried out.2.In this paper,principal component analysis(PCA)was carried out on the hyperspectral images of blood stains in the band range of 380 nm-1000 nm,and 500 nm-1000 nm was selected as the best spectral detection area for blood stains recognition according to the spectral absorption characteristics of blood.Based on the spectral imaging characteristics of the best spectral region and principal component,three characteristic bands are selected in this paper.A mixed 2D-3D convolutional neural network(2D-3D CNN)model for blood pattern detection and location recognition was designed by calculating the average spectrum of the blood stain region and combining with spectral imaging processing operation.Through the detection of 103 blood and blood analogue samples on different substrates,the average recognition rate was 95.39%.3.In this paper,spectral imaging technique is used to obtain the mean spectral curves of blood stains within a range of pixels.Three characteristic bands 410 nm-430 nm,540nm-580 nm and 800 nm-1000 nm were selected based on the deformation characteristics of hemoglobin and water in the bare environment.With the support of statistics and weight analysis,a mathematical function mapping model was built for the exposure time of blood and spectral reflection intensity,and the prediction accuracy of this model was 94.6%.At the same time,this method is also compared with the current more advanced blood stain exposure time prediction methods(such as: support vector machine-SVM,KNN,genetic interval partial least square method,etc.).The experimental results show that this method has more advantages in the prediction of blood stain exposure time within 30 days. |
PDF Full Text Request