Clustering of fish freshness using discrete wavelet transform and Kohonen self organizing map

Fish freshness is one of the criteria of fish, whether it is good or not for consumption. A fish can be categorized as a fresh condition if it looks like alive fish. The observation about level of fish freshness generally can be checked directly through the human senses, but along with the development of technology, fish freshness observation can detected using a technology. One of them is by utilizing digital image processing. The purpose of this research is to cluster fish into three clusters namely fresh, not fresh, and rotten using the discrete wavelet transformation and Kohonen Self Organizing Map (SOM). Stages of clustering of the fish freshness are pre-processing, feature extraction, and clustering. At the pre-processing stage, RGB to grayscale images are converted. After pre-processing, the next stage is image decomposition in pre-processing result which is applied using Discrete Wavelet Transform (DWT) Haar level 3 and takes the statistical parameters of the mean and standard deviation from the horizontal detail coefficient. Mean and standard deviation obtained are used as the input in clustering process using Kohonen SOM. Based on the test result, it showed that the percentage of fish was clustered correctly is 92,857%.