Automatic Identification Algorithm And Software For Gem Species Based On Multi-spectral Data Analysis

Gemstones are the mainstream products in the jewelry and jade industry and are deeply loved by people.Effective identification of gemstone species is of great significance to safeguarding consumers' rights and interests.As a non-destructive testing method,spectral analysis provides a new research method for identifying natural diamonds and synthetic diamonds,and is widely used in the field of jewelry identification.The rapid development of automation technology and artificial intelligence technology has promoted the digital,networked,and intelligent transformation of jewelry identification.While improving the accuracy,repeatability and efficiency of identification,it reduces the requirements for labor.Focusing on the purpose of automatic,efficient and accurate analysis and identification of gem species,this paper researches and develops gem species identification algorithms and software based on multispectral data analysis to achieve high-precision and high-efficiency identification of gem species.The specific research content and work are as follows:1.According to the characteristics of different spectral data,an adaptive preprocessing method is proposed.For spectra with severe glitch noise,a Savitzky-Golay smoothing filtering method with adaptive optimization parameters is proposed,which can better maintain the shape of the original spectral curve while filtering out the noise.For spectra with inconsistent dimensions,the reasons for the inconsistency of dimensions in different spectra were analyzed,and their respective dimensional correction methods were proposed.For spectra with large differences in absorbance and transmittance,the min-max standardization method is used to eliminate the influence of singular sample data.2.A method of infrared reflectance spectroscopy combining characteristic peak analysis and spectral matching is proposed to realize the identification of gemstone materials.For the problem of the offset of the characteristic peak of the spectral curve,based on the central limit theorem and the 3? principle,the population is estimated with samples,and the corrected 3? offset interval of each characteristic peak of each infrared reflectance spectrum is determined.For further comparison,the Pearson spectrum matching algorithm is used to realize the accurate distinction of gem material and its derived minerals.The experiment verifies that the algorithm combining characteristic peak analysis and spectrum matching has high reliability and stability.3.Aiming at the problems of artificial processing of gemstone infrared transmission spectrum data and ultraviolet-visible spectrum data sequence is long,the peak waveform difference is large,and the characteristic peak is not obvious,a complex spectrum identification based on the N VS 1 model structure of the double-layer LSTM network is proposed.Method to achieve high precision identification of artificially processed gemstones.Experiments verify that the model has good robustness and recognition ability,and can achieve the purpose of accurately identifying whether the gemstone has been manually processed.4.On the basis of the above-mentioned key technology research results,a gem species identification software based on multispectral data has been developed,and a large number of experimental tests have been carried out to verify the accuracy,stability and applicability of the algorithm and software.