Authenticity Detection for Mechanical Watch Using Hidden Markov Model

The detection of original products is needed by a brand to maintain the authenticity of goods. This is to prevent goods from being imitated by ensuring the quality of goods to users. The issue presented in this study is how to use HMM to determine the authenticity of watch. At this stage of determining the authenticity of watches, the HMM algorithm is used to look for color matching, tough glass, and the presence of mechanical watch hands. Authenticity standards are carried out using probabilities including emission probability and HMM transition probability. However, before entering processing using the HMM algorithm, the system needs to go through the pre-processing and data training stages to be able to train the system to process mechanical hour data. The thing that influences this research is the distance between the photo object and the camera in the range of $5-20 \mathrm{~cm}$ and not too blurry or vibrating. The results of the research will be in the form of a statement that mechanical watch fall into the Bad, Good, or Tolerance categories. Watches that have a Bad category statement mean they are not fit for sale; Good means worth selling; while Tolerance means it requires manual re-detection. System testing using the black box method in this test obtained an accuracy value of $88 \%$ of the total data testing of 100 mechanical clock images.