The Non-destructive Detecting Technology Research Of The Sugar Content Like Navel Oranges Of Near Infrared Spectral

This paper use the like navel oranges of gannan as the research object and simply introduce the basic principle of the near infrared spectral(NIR) analysis technology, process and the implementation steps, influence factors and the detection method. Searching many reports of the near infrared spectral analysis technology of agricultural products of the internal quality nondestructive testing progress in the currently during the domestic and foreign, learned about where the near infrared analysis technology had been used. This paper mainly includes:the effect factors of the test accuracy, tested some kinds of spectral measurement method for this test and finally get the method suitable for this experiment, compared several different spectral pretreatment methods and briefly introduces the method of optimal wavelength and several kinds of modeling method to be used in this experiment. The main content of this study and the conclusion are as follow:(1) Simply analysis the spectroscopy technology and the principle of this technology used in sugar content detection, knowing something about the industry in the domestic and international dynamic, and determine the actual needs of this experiment.(2) Introduce the overall design and the software and hardware the experiment used, including:Spectrometer, small units, samples picking and storage, spectrum acquisition software, data processing software, modeling software of introduction, and how to use these test equipment.(3) Based on the results of other researchers and combined with the test their own conditions to study the experiment, we select some factors that may influence the accuracy of experiment,such as:temperature, the size of samples, the way of the sample, the strength of the light source, and finally form a set of scheme suitable for this test.(4) This chapter introduces the spectral pretreatment method, the optimal wavelength selection method, modeling methods, such as:differential method, convolution smoothing, wavelet transform method, the interval partial least square method, and use these methods to this experiment, the last,compared the actual effect of various methods, Finally, draw out that use the original spectrum data that not had pretreatment is the best fit for the experiment, and concludes that the correction set r2=0.8191, RMSEC=0.4301, forecast set r1=0.7544, RMSEP=0.4931. Has relatively high r2, lower RMSEC and RMSEP, and calibration set r reached0.9.(5) At the last chapter, to sum up what things the experiment had done, explain some hortcomings of the experimental may have,and put forward some suggestions for the future work.