Design And Implementation Of An Embedded High Spectrum Resolution Video Capture System

Compared with RGB color camera, the traditional spectrometer is enable to obtain more color information, which can be used in many fields, such as military, medical, material detection and scientific observation. But the traditional spectrometer mostly adopts the principle of sequential exposure or spatial scanning, at the expense of time information or spatial information for the spectral informatfon, so that we can not get high spectral change the information in the dimension of time or space, usually is used for static observation of samples. Because of its custom mechanical structure and special optical elements, large volume and expensive price, the application of the traditional spectrometer is limited. High spectral camera based on RGB and gray binocular symmetry, makes fully use of the high spatial resolution of RGB pixel value and high spectral resolution gray scale camera response value, executing the bilateral filtering algorithm to achieve high spatial and high spectral resolution in real time Acquisition. It compared with the traditional spectrometer in structure principle and effect is improved obviously. But this high spectral camera using the X86 computers as its control center, has a evident defect which is its large volume of high power consumption and inconveniently to be carried out. At the same time the bilateral filtering algorithm in CPU execution with slow speed, also make limits for its application.For the problems of high power consumption and large volume of the hyperspectral camera, this paper combines embedded system technology with prism mask type hyperspectral data acquisition theory, realizes an embedded hyperspectral and high spatial resolution video acquisition system. Based on Tegra K1 processor with low power consumption connected to the front cameras and the back display device, using the Tegra K1 processor peripheral scalability, the system of acquisition and processing display module are integrated to reduce system size, reduce the power consumption of the system. At the same time by using Linux kernel based embedded Ubuntu system with Qt graphic library, the application of interactive design, through the control button on the interface to complete the acquisition of RGB and gray of binocular camera video images, high resolution spectral data, any pixel spectral curve and pseudo color image Synthetic display and other functions, convenient operation and use, greatly improve the practicality.For the problem of the algorithm on the CPU execution speed slowly, this paper use GPU CUDA core of Tegra K1 processor to realize in CUDA multi core parallel implementation of bilateral filtering algorithm to obtain the hyperspectral data process, improved the algorithm execution speed, than the x86 CPU of the computer algorithm execution speed is increased more than 5 times, compared with Tegra K1 CPU algorithm execution speed is improved by more than 10 times, the performance has improved significantly.