Research On Spectral Data Processing System Based On Linear CCD

With the development and progress of science and technology, the spectral analysis which is an important way to analysis material composition and characteristics, has been increasingly wide application. Given that the traditional spectrometer can not detect multiple wavelengths of light at the same time, besides, its large volume, complex structure and higher using requirement restrict its promotion speed. The micro spectrometer based on CCD has become a fastest development branch in spectrum instrument.In this paper, through analysis the working principle of CCD, the author choose a linear array CCD as spectrum detection device, design a high-speed spectral acquisition and processing system, achieve the spectral data pre-processing, study the spectral data wavelength calibration, and completed the spectral data processing software.Spectral data processing system can be divided into four parts: the spectral acquisition part, de-noise, wavelength calibration and software design part. Spectral data acquisition part is mainly based on ILX554 B linear CCD as photoelectric converter, using C8051F500 microcontroller as the control core, using CPLD generated the linear CCD timing drive signal, AD9220 ADC signal and the correct timing of buffer IDT7203 to write signal, so that this circuit achieved synchronization. According to the spectral data containing noise features that are Gauss white noise and pulse noise, a combination filter based on the wavelet threshold is presented, which is employed to reduce the spectral data noise. The combination filter principle that consists of wavelet threshold method and median filter is mainly introduced, and the simulation experiments were carried out. The simulation results show that using the combination filter method, in which the wavelet threshold method is used after the median filter, the noise of the spectral data can be effectively reduced. The method adopted symmetrical zero area methods of peak-detection algorithm and the least square fitting, and chose symmetrical square wave function as a window function to make a convolution with mercury lamp spectrum to find the peak point. Finally, this paper used fourth-order method of least square fitting algorithm for wavelength calibration. Experimental results show that the precision of this algorithm was improved in some extent; using C# prepared spectrum data processing software system to control the spectral acquisition system and finish spectral de-noise, peak processing, spectral calibration and other basic functions.