Research On Material Detection And Identification Based On Spectral Imaging

Spectral imaging detection is a new type of photoelectric detection technology,which refined its image in spectral dimension on the basis of the traditional RGB image,and has a wide range of applications in the field of digital image processing.Spectral images are divided into multi-spectral images and hyper-spectral images,between which obvious differences in spectral resolution and band are observed.Traditional multi-spectral images generally consist of 3 to 10 bands,while hyper-spectral images can be composed of tens to hundreds of bands,with resolutions up to nanometers level.Common hyper-spectral images have an FWHM of 5-10 nm.With rich spectral information,spectral images have been widely used in medical,security,military,geology and agriculture field in recent years.Currently,with the rapid development of military technology,information acquisition has become the key factor in warfare.Spectral imaging technology can detect the physical indicators and chemical components of the target,and can position target and detect material through qualitative and quantitative analysis,which is of vital importance for handling the ever-changing battlefield environment.In this thesis,based on actual imaging requirements and application tasks,a set of complex weather spectral imaging system based on traditional rotating filter wheel was developed for material detection and identification.This system designed the complex weather environment,dark room,imaging module,data processing module of imaging system.Subsequently,based on the rotating filter wheel system,a set of dual-camera spectral imaging system was developed to simulate the environment under complex weather conditions,including foggy and cloudy days.Spectral data of military models with different materials were obtained.In this thesis,spectral image pre-processing methods such as contrast lifting technique under haze weather based on atmospheric scattering proximity effect,and Retinex-algorithm-based dark scene image enhancement were used to reduce the loss of spectral information.According to the obtained discrete spectral curves,an interpolation-based spectral extension algorithm was proposed.The spectral database of military models made of different materials such as alloy,plastic,wood and iron was constructed in advance,and the spectral images of these aircraft and tanks were obtained during the experiment.A variety of spectral matching algorithms for material identification testing,simulation and practical experiments showed that this system and method can effectively detect and identify properties of materials.