One of the most important selective specifica-tion in determining the quality of fish is its freshness. In this study, the fast and sensitive alternative tech-nique is proposed as the fish freshness detection.120 different images of 20 red mullet fish images were taken with an interval of 8 hours in two days. The segmentation of fisheyes according to a fully auto-matic computerized vision was acquired from the digital images of fish samples. Three different meth-ods have been proposed to identify the fish freshness after the segmentation process. The first method is the local binary pattern(LBP) method. The distance between histograms obtained by this method was ex-amined and compared with Chi-Square Divergence, Kullback-Leibler Divergence,and Correlation meth-ods. Next, the discrete wavelet transform was per-formed. The energy densities of these wavelet trans-forms were investigated. The data obtained from these methods provide a strategic framework for the assessment of fish freshness. Then, the histograms obtained from the LBP method were educated with supervised machine learning. As a result of the tests, the freshness situation of 18 fish was determined within 20 red mullet fish.
展开▼
