Monitoring of fresh-cut spinach leaves through a multispectral vision system

This paper reports the development of image processing methods for the detection of superficial changes related to quality deterioration in ready-to-use (RTU) leafy spinach during storage. The experiment was performed on spinach leaves stored at 4.5 °C for 21 days (Set 1) and at 10 °C for 9 days (Set 2). Regarding Set 1, 75 units were evaluated beginning at time zero and after 7, 14, and 21 days of storage (treatments t1.0, t1.1, t1.2, and t1.3, respectively). In the case of Set 2, 24 samples were measured at time zero and after 3, 6, and 9 days (treatments t2.0, t2.1, t2.2, and t2.3, respectively). Multispectral images were acquired using a 3-CCD camera centered at the infrared (IR), red (R), and blue (B) wavelengths. Opportune combinations of these bands were calculated using virtual images, and a non-supervised classification was performed. A large number of spinach leaves belonging to Set 2 showed injuries due to the effects of in-pack condensation; thus, an image algorithm was developed to separate these defective leaves before applying the classification. For Set 1, Set 2 and all the calculated virtual images, the classification procedure yielded two image-based deterioration reference classes (DRCs): Class A, including the majority of the samples belonging to t1.0 and t1.1 (Set 1) and to t2.0 and t2.1 (Set 2) treatments and Class B, which comprised mainly the samples belonging to t1.2 and t1.3 (Set 1) and to t2.2 and t2.3 (Set 2) treatments. An internal validation was performed, and the best classification was obtained with the virtual images based on R and B bands. For each sample, camera classification was evaluated according to reference measurements (visible (VIS) reflectance spectra and CIE L*a*b* coordinates); in all cases, VIS reflectance values corresponded well with storage days, and Classes A and B could be considered homogenous with regards to L* and a* values. Taken together, these results confirmed that a vision system based on R and B spectral ranges could constitute an easy and fast method to detect deteriorating RTU packed spinach leaves under different refrigeration conditions.

► An imaging process to classify RTU spinach leaves according to color degradation during storage was proposed. ► Leaves with injuries due to in-pack condensation were separated through an image algorithm. ► The image analysis was based on the relative histograms of four virtual images (R/IR, IR − R/IR + R, B/R, and R − B/R + B). ► During the classification process two classes were identified, comprising the majority of the samples analyzed on the first and on the last days of storage. ► Virtual images based on R and B bands gave better validation results than those based on IR and R ones.