Modification of the Haar Wavelet Algorithm for Texture Identification of Types of Meat Using Machine Learning

This paper proposes a modified Haar wavelet algorithm to identify the texture of various types of meat using machine learning (ML). In Indonesia, ahead of Eid al-Fitr, the price of beef continues to increase, of course causing a decrease in beef sales. To anticipate this, some traders mix beef with pork. The pork was chosen because the price of pork is cheaper and the color and texture of pork are similar to beef. In plain view, beef and pork are difficult to distinguish for the layman. Therefore, it is necessary to have a system that can distinguish various types of meat. The samples used in this study used the texture of beef, buffalo meat, mutton, horse meat, and pork. Image processing is done by calculating the Red, Green, and Blue (RGB) values, Gray Level Co-occurrence Matrix (GLCM), and Hue, Saturation, and Value (HSV) for each meat image, then normalizing is done to get RGB, GLCM values, and HSV. The value generated from image processing is used as input for the parameter verification program using machine learning. Then after getting the values of these parameters, the values will be converted into Haar wavelets and modified Haar wavelets using machine learning. This paper aims to examine the performance of Haar wavelets and modified Haar wavelets in identifying the texture of various types of meat using machine learning. The lowest accuracy on the value of mutton is 76.72% which is identified as a Wavelet Haar, while the highest accuracy is on the value of beef which is 153.45% which is identified as a modified Wavelet Haar. Among the various methods used, the Haar wavelet modification produces the highest accuracy of 153.45% on the value of beef compared to the Haar wavelet method which produces the lowest accuracy of 76.72% on the value of mutton.