A Nondestructive On-Line Detection System of Apple Internal and External Parameters Based on Optical Technique

Long Li,Yankun Peng,Yongyu Li,Yachao Liu

DOI: https://doi.org/10.13031/aim.201900783

2019-01-01

Abstract:On-line detection technology is an important means for rapid determination of internal and external quality of apples after harvest. Generally speaking, the customers' first purchase usually depends on the fruit‘s appearance rather than its price. However, its internal quality determines whether or not it will be bought again by the same group of consumers. In order to address the market demand for high-quality apples, we developed an on-line optical and spectroscopy system to examine simultaneously both the internal and the external quality of the samples. In terms of external quality, it mainly refers to the size and the bruise of apples. Besides, the internal quality mainly includes its soluble solid contents. The on-line detection system is composed of conveyor belts, the external quality detecting mechanism and the internal quality detecting mechanism. The conveyor belts transferred an apple to the external acquisition device, and the rotation completed when the sample moved forward. An automatic segmentation and combination method was developed based on the Open CV library to determine the synthesized image, which contains the whole surface information of the apple. In order to solving the problem of uneven brightness distribution of apple images, a novel image segmentation method was developed on the basis of the isohypse line in the raw image. In fact, the results revealed that the correct rate of external quality assessment was 96.76%, the correlation coefficient of the size detection was 0.9763, and the root-mean-square error was 1.3243 mm. Furthermore, the apple was then shifted to the internal quality detecting mechanism which was illuminated by two 10W tungsten halogen lamps. The range of spectrum from 400nm to 1100nm was detected by the spectrometer, and the linear relationship between the spectrum and the measured values was delineated with the PLS method. Three methods, including the multiplicative scatter correction (MSC), standard normal variable correction (SNV) and optical path length estimation and correction (OPLECm), were adopted to eliminate the effect of light scattering as much as possible in the spectrum in comparison with the differences. The correlation coefficients involving the calibration and prediction of SSC were 0.9292 and 0.9219; and the root-mean-square errors with regard to the calibration and the prediction were 0.5482% and 0.5867%, according to the OPLECm method. The detection result of the external quality and the modeling result of the internal quality indicated that this system could effectively pick out high-quality apples. In other words, this on-line optical and spectroscopy system could indeed be utilized for the quick and on-line quality assessment of apples.