Study Of Airborne Spectral Video Acquisition System And Algorithm

Spectral imaging technology combines with the advantages of spectral technology and traditional imaging technology.And it can obtain responses of light in different wavelength range so that the 2D imaging data and the 1D spectral data of objective can be acquired.Thus the spectral imaging technology can be applied to broad fields,such as military investigation,material identification and medical diagnosis etc.Traditional spectral imaging devices are costly as well as large in volume.And it can hardly capture dynamic spectral information due to the adoption of time-series scanning method.Acquisition of dynamic spectral video,miniaturization and mobility of spectroscopy devices are key factors to promote spectral imaging technology to be further widely applied.This paper focuses on the imaging principle and the accelerated spectral reconstruction algorithm of the airborne spectral video acquisition system,and a multi-spectral face liveness detection algorithm.The main work and innovations are described as follows:1.This paper develops a miniaturized airborne spectral video acquisition system.We design the architecture of the system.And there is also a spectral slice compression algorithm proposed,which greatly improves the compression ratio of spectral data and reduces the transmission bandwidth of spectral data.Thus the system can capture and process spectral data in real time with spectral resolution of 2-6 nm,spatial resolution of 2048 × 1536,and video frame rate of 20 fps.2.This paper introduces a spectral reconstruction algorithm based on CUDA acceleration.Spectral reconstruction algorithm realizes the parallel calculation by use of the hardware and software architecture features of CUDA,which improves the spectral reconstruction speed by about 50?75 times.Thus it realizes the real-time reconstruction of spectral data and expands the application of spectral system.3-This paper proposes a face liveness detection algorithm based on multi-spectral technology.It utilizes the spectral characteristics of facial skin so that it can improve the accuracy and stability of face liveness detection.Additionally,the proposed algorithm can defend common fraudulent in face recognition.