บทคัดย่องานวิจัย

หมวดหมู่ : ฐานข้อมูลงานวิจัย

Image classification of bananas (Musa cavendish) during ripening based on appearance features

Mandoza F. and Aguilera J.M.

5th International Postharvest Symposium . Volume of Abstract . Verona, Italy 6-11 June 2004, p.117

2004

บทคัดย่อ

Image classification of bananas (Musa cavendish) during ripening based on appearance features     In the trade, the ripeness of bananas is normally assessed visually by comparing the colour of the peel to standardized colour charts that describe seven ripening stages, and sometimes by instrumental techniques (i.e., colorimeters). Human visual inspection is a highly subjective and tedious process. In contrast, colorimeters allow accurate and reproducible measurements of the colours. However, the surface colour must be homogeneous and many locations must be measured to obtain a representative colour profile. In addition, in the case of bananas these techniques are usually destructive requiring the removal of the peel for the measurement.

The objectives of this study were: (i) To implement a computer vision system to characterize quantitatively colour changes during ripening of bananas; (ii) To identify features of interest which can be related with the later ripening stages such as development of brown spots and textural features of the images; and (iii) To develop a statistical model to identify the ripening stages of bananas from samples previously classified by expert visual inspection. Nine simple features of appearance: L*a*b* values, brown spots area percentage (%BSA), number of brown spots cm-2 (NBS cm-2) and homogeneity, contrast, correlation and entropy of image texture were evaluated for classification purposes.

Results show that parameters a*, %BSA, NBS cm-2, and contrast better depicted the appearance characteristics as an indicator of banana ripeness. Thus, in spite of the inherent variability of banana samples the proposed computer vision technique combined with a simple classification technique has great potential to differentiate among ripening stages of bananas as professional visual perception. Using L*a*b* bands, %BSA and contrast it was possible to classify 49 banana samples in their seven ripening stages with an accuracy of 98%. Computer vision shows promise for o­n-line prediction of ripening stages of banana.