Fruit Quality Grading Using Texture Feature-Based PLS-DA Technique

Abstract

During the postharvesting of horticulture, the grading of the fruit is significant because it determines the satisfaction and preference of the consumers while reaching the market. The fruit grading using the physical classification is an expensive task and inaccurate classification may occur due to human errors. Thus, there is a need for automatic fruit grading using the non-destructive process. This research proposes an efficient fruit grading technique using the partial least squares-discriminant analysis (PLS-DA technique) based on the texture features. Here, the features such as local binary pattern (LBP), local directional pattern (LDP), local optimal oriented pattern (LOOP), local gradient pattern (LGP), and Local Ternary Pattern (LTP) are extracted for the classification of the fruit. The feature extraction based on these texture features is employed after pre-processing the multi-spectral input image. From, the extracted features, the classification of fruit such as apple, banana, pomegranate, and mango are employed. Then, the parameters such as firmness, soluble solids concentration (SSC), and titratable acidity (TAC) are evaluated for the fruit quality grading using the PLS-DA technique. The proposed method is evaluated in terms of accuracy, error, [Formula: see text], residual predictive deviation (RPD), sensitivity, and specificity and obtained the values of 95.67%, 4.33, 92.97%, 0.05, 95.66%, and 95.17% respectively.