Pork quality assessment through image segmentation and support vector machine implementation

Pork is the most consumed meat in the Philippines, and efficient quality control is essential for ensuring the safety of its consumers. Current manual procedures of meat inspection are time-consuming and laboratory-intensive considering the large amount of supply to be examined. This research aims to construct a rapid objective system of pork quality assessment with respect to meat freshness through Support Vector Machine (SVM) implementation, and to ultimately have an accuracy rate of ? 90%. 35 meat samples were collected, and their images were acquired. 30 of these were randomly designated as part of the training dataset while the rest were designated as part of the testing dataset. Of the 30 training samples, 6 were randomly chosen for the creation of a microbial profile. In all of the acquired image samples, image segmentation was performed and the RGB, HSV, Lab, and statistical texture features were extracted. These were inputted in 15 different SVM configurations. SVM classification yielded an accuracy rate of 93.33 %. Results from the microbial profile revealed considerable microbial activity at the 5th and 6th intervals (10th and 12th hour) with 2 and 3 colonies formed, respectively. With the ability of the SVM to distinguish between samples with respect to the hour interval and with the supplementation of the microbial profile, an objective artificial intelligence mechanism for freshness detection was successfully created