The pineapple maturity level also referred as pineapple maturity index is based on the percentage of yellowish that appears on the pineapple’s skin. In pineapple industry to determine the level of maturity, human experts adopt methods based on their subjective assessment of skin color. To this day, the pineapple maturity sorting process is still performed manually by expert human grader. So in order to reduce errors caused by human factors, there is a need to automate this process to an automated inspection system. The matured fruit harvested for the purposes of local sale or export is complete fruit with crown, fruit body and stump. However, in determining the pineapple maturity index, the main thing to be considered is only the pineapple fruit without crown. Fruit without crown also represents the actual size of the pineapple. Therefore the percentage of yellowish must be proportional to the size of the pineapple. Having extensive search of literatures found that studies of the size of the fruit, especially pineapple are very limited and only been started in recent years. To obtain the actual size of the fruit, the detection Region of Interest (ROI) is using segmentation method called minimum symmetrical edge distance. This minimum symmetrical edge distance algorithm wills geometrical rotated the pineapple images which to align with horizontal axis. Then the shortest vertical distances of the edge is calculated and converted to a background pixel, the largest region (fruit body) is maintained and the small region (crown) was abolished. The performance of segmentation algorithms are calculated using misclassification error that provides the rate of image pixels are incorrectly misclassified into the wrong segment. The results reveal that the algorithm used to achieve overall accuracy up to 99.05%. ROI that has been identified lengthened for feature extraction on the skin color of pineapple. Statistical based features namely minimum, maximum, arithmetic average and standard deviation were extracted from each image channels within detected ROI to represent pineapple skin color's tendency and dispersion. Next, classification index to determine the pineapple maturity level has been applied which are linear classification using thresholding value and artificial neural network adopting pattern recognition method. The results show that the classification using artificial neural network (pattern recognition) involving feature vectors arithmetic average and standard deviation for all channels R, G and B give the average correct classification rate of 88.89%.
展开▼
