<![CDATA[This study was conducted to examine the effect of grayscale image variations on the accuracy of fish species recognition by utilizing fish scale images through the Artificial Neural Network (ANN) method. Automatic fish species identification plays a crucial role in the fisheries sector, both for research purposes, marine resource monitoring, and trade processes. One factor that can influence recognition accuracy is the quality of image representation, including the grayscale level used. Therefore, this study aims to analyze how much grayscale level variations affect fish species classification results. This research method uses a dataset consisting of 180 scale images for each fish species. Of these, 150 images are used as training data and 30 images as test data. The feature extraction process is carried out using the Gray Level Co-occurrence Matrix (GLCM) method, which utilizes contrast, energy, homogeneity, correlation, and entropy parameters. These features are then used as input to the ANN for the classification process. The analysis was conducted by comparing the accuracy results of various grayscale levels, namely 16, 32, 64, 128, and 256 levels. The results showed that variations in grayscale significantly influenced the accuracy level of fish species recognition. The highest accuracy was obtained at a scale of 256 levels with a value of 96%, followed by a scale of 128 levels at 95%, 64 levels at 92.5%, 32 levels at 84.2%, and the lowest at 16 levels with an accuracy of only 82.5%. In conclusion, the higher the variation in grayscale levels used, the better the recognition accuracy obtained. Thus, the use of images with 256 grayscale levels is recommended for research on fish scale image classification using the ANN method because it is able to provide the most optimal results.]]>
