The Application Of Hyperspectra Technology In The Detection Of Apple

During the process of picking or transportation, the apple may be damaged by external forces. The internal quality of damaged part has changed but it is not obvious, so it is important to detect the external damage of apples. The consumers not only care about the external quality of fruit, but also pay great attention to the internal quality of fruit at the same time.This paper chooses the Fuji apple as the research object and studies on the apple’s external injury by using hyperspectral imaging technology(380-1038nm). The whole band of apple image was detected by using the principal component analysis method and selected the principal component image. Then, according to the feature vector of the principal component image, the 10 characteristic bands were selected, and then the principal component analysis was done for the characteristic band, the fourth principal component images(PC-4) were selected to do the image processing and recognition, and the recognition accuracy was only 81%, which was due to the affects of the spot. In this paper, the image difference algorithm was used to eliminate the influence of spot and the recognition rate increased to 90%. Secondly, in this paper, using hyperspectral technology nondestructive to detect the sugar content and p H value of apple. The algorithm of Multiplicative Scatter Correction spectral pretreatment(MSC), Savitzky Golay(S-G) convolution smoothing and MSC+S-G were used to deal with the original spectrum, then modeled the full band spectral modeling by using the Partial Least Squares Regression(PLSR) and Principal Component Regression(PCR). According to the correlation coefficient method to select the characteristic band, using partial least squares regression(PLSR) and principal component regression(PCR) to establish the prediction model. The results showed that,the PCR model had the best effect after S-G convolution smoothing pretreatment in terms of the Brix value, the average relative error of the real and predictive value was 0.022877. The PCR model had the best effect after MSC pretreatment in terms of the pH value, the average relative error of the real and predictive value was 0.014614. Finally, the paper put forward a research method using hyperspectral technology for the detection of apple taste(sourness and astringent aftertaste). Using the correlation coefficient method to select the characteristic wavelength and established the PLSR prediction model of the taste index under the characteristic wave band. The results showed that, it is feasible to detect the taste index of apple by using hyperspectral technology. In the prediction model of sourness, the correlation coefficient was 0.9700,the root mean square error of prediction was 0.8587 and the average relative error was 0.042189. In the prediction model of astringent aftertaste, the correlation coefficient was 0.9115,the root mean square error of prediction was 0.0843 and the average relative error was 0.146391.In summary, hyperspectral imaging technology not only can effectively detect the apple external damage, but also on the apple internal quality index and taste index, and it provided experimental basis for the subsequent online detection.