| dc.description.abstract | Sweet orange fruit grading in Indonesia still done manually. Manual grading has many shortcomings, such as require a relatively long time, produce a diverse sorting, and differences in the quality perception of products as an element of subjectivity. Based on this, it needs a method that can classify the quality of sweet orange fruit effectively and efficiently. Image processing is an alternative to overcome this. The purpose of research are: (1) determining and analyzing the variables of image quality sweet oranges to construct algorithms sweet oranges grading, (2) creating a model of logic equations to algorithms sweet orange grading, and (3) determining sweet citrus fruit grading validation program based on variables and logic equations which have been found. The results is expected in the process of sorting produce quality fruit sweet orange uniform at every grade quality and sorting is done is not damaging materials (non-destructive). The sample used in this study are the sweet orange fruits (sukarri varieties) obtained from Junrejo Village, District Junrejo, Batu Regency. Materials used in this study are sweet orange fruit with S (super) quality classes, A, B, and RJ (reject) each 50 samples for data modeling logic equations, and 15 samples for validation data from each class quality, so that the whole sample is 260 pieces. After taking the image of the sweet orange fruit using a CCD camera, then conducting image analysis to obtain seven image quality variables ie: seven area, height, width, perimeter, area defects, color index r, and color index g, using image processing program. Seven variables have been obtained from the image processing program with pixel units, will be analysis statistically using Boxplot. Statistical analysis intending to find the range of image quality variables as the number of input logic sentences. From the seven variables of image quality obtained four image aquality variables matching to quality criteria, namely the area (a), defect area (ac), green color index (g), and color index red (r). Limit values that can separate quality Super, A, B, and Reject based on image quality variables are, for the super quality a ≤ 278602, a ≥ 147125, c ≥ 204, c ≤ 11551, g ≥ 0,58, g ≤≤ 0,66, r ≥ 0,27, and r ≤ 0,41, for A quality, a ≤ 146733, a > 138120, c ≥ 204, c ≤ 11551, g ≥ 0,58, g ≤ 0,66, r ≥ 0,27, and r ≤0,41, for B quality, a ≤ 138985, a ≥ 107261, c ≥ 204, cc < 11551, g ≥ 0,58, g ≤ 0,66, r ≥ 0,27, and r ≤ 0,41, as for the Reject quality: a ≤ 109994, a ≥ 55443, c ≥ 3056, c ≤ 42029 or c ≥ 204, c ≤ 11551, g ≥ 0,66, g ≤ 1,55, r ≥ 0,41, and r ≤128. Based on the limit values above, then obtaining 9 combination models that can separate the logic equation between quality S, A, B, and RJ. The validation process of sweet orange grading program has total acurracy 85%. | en_US |
