Identification of the freshness level of Tuna based on Discrete Cosine Transform on Feature Extraction of Gray Level Co-Occurrence Matrix using K-Nearest Neighbor

Abstract

Gorontalo Province is one of the provinces that have fishery potential and has a large sea area that can be managed to support the economy and development of the province. Gorontalo is also one of the tuna-producing provinces in Indonesia, where tuna is also one of the mainstay fisheries commodities. This study aimed to combine transformation and texture feature extraction methods to improve the identification of the freshness level of tuna. This research used Discrete Cosine Transform as transformation detection and Gray Level Co-Occurrence Matrix as texture feature extraction. To find out the value of the proximity of the training data and image testing of tuna fish, the K-Nearest Neighbor classification method was employed. Then, the Confusion Matrix was used to calculate the accuracy level of the K-Nearest Neighbor classification. This research was carried out with 4 stages of testing, namely at angles of 0°, 45°, 90°, and 135°, and using the values of k=1, 3, 5, and 7. The test results of using training data of 428 images and testing data of 161 images in four classes used with angles of 0°, 45°, 90°, 135°, and the value of k=1, 3, 5, 7. The highest accuracy results was obtained at an angle of 0° with a value of k = 1 of 94.40%, while the lowest accuracy value was at an angle of 90° and 135° with a value of k=7 of 59%. This showed that the Discrete Cosine Transform transformation method was very effective to improve the performance of texture feature extraction of Gray Level Co-Occurrence Matrix in extracting tuna image features. It was proven from the results of the accuracy of the K-Nearest Neighbor classification obtained.

Keywords

• identification
• classification
• dct
• glcm
• knn
• confusion matrix