Submitted to: Postharvest Biology and Technology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: April 1, 2003
Publication Date: February 1, 2004
Citation: Lu, R. 2004. Multispectral imaging for predicting firmness and soluble solids content of apple fruit. Postharvest Biology and Technology. 31(2):147-157. Interpretive Summary: Currently, machine vision is widely used in the fruit industry for sorting fruit for color and size. However, sorting fresh fruit for texture and flavor is a challenging issue, which would provide the industry with an essential means for assuring the quality and consistency of the fresh product delivered to the consumer. The object of this research was to investigate the usefulness of a novel optical technique using multispectral imaging for nondestructive sensing of firmness and sweetness of apple fruit, two important quality attributes that directly influence consumer purchase decisions. Multispectral imaging is a technique that acquires multiple images at selected wavelengths that are useful for revealing quality attributes of fruit. A multispectral imaging system was developed to acquire images at five wavelength bands in the spectral region longer than the visible light. Computer algorithms were developed to extract useful features from these images for predicting fruit firmness and sweetness. The multispectral imaging system was able to predict fruit firmness, in respect to the standard destructive firmness measurement method, with the coefficient of correlation of 0.87 and the prediction error of 5.8 N, less than a minimum firmness value of 6.0 N a trained panelist could detect. The system also gave relatively good predictions of fruit sweetness with a correlation of 0.70 and the prediction error of 0.91 Brix. The multispectral imaging technique is promising for grading and sorting fruit for firmness and sweetness. With further improvement in system hardware and imaging speed, the technology can help the fruit industry in delivering better and more consistent quality fruit to the consumers and improve the competitive advantages and profitability of the U.S. fruit industry.
Technical Abstract: Firmness and sugar content are important consumer quality attributes for apples and many other fresh fruits. This research investigated the feasibility of using multispectral imaging to quantify light scattering profiles from apple fruit for predicting firmness and sugar content. Scattering images, generated at the fruit surface from a broadband, focused beam (0.8 mm diameter), were obtained from Red Delicious apples for five selected spectral bands (10 nm bandpass) between 680 nm and 1060 nm. Ratios of scattering profiles for different spectral bands were used as inputs to a backpropagation neural network with one hidden layer to predict fruit firmness and sugar content. The three ratio combinations with four wavelengths (680 nm, 880 nm, 905 nm, and 940 nm) gave the best predictions of fruit firmness, with r = 0.87 and standard error of prediction (SEP) = 5.8 N. Only one ratio involving 905 nm and 940 nm was needed for predicting the sugar content of apples with r =0.70 and SEP = 0.91%. The multispectral imaging technique is promising for grading and sorting of fruit for firmness and sugar content.