Fish cultivation is a vital economic activity for coastal communities, yet traditional farming methods often face challenges such as environmental instability, feeding inefficiencies, and water pollution. Effective monitoring of underwater environments is essential to improve fish quality and farming efficiency. A crucial part of this process is the accurate classification of fish and non-fish objects. This study proposes a method for underwater classification using morphometric feature extraction and machine learning techniques. The research process involves six main steps: (1) preparation of Region of Interest (ROI) detection data, (2) extraction of morphometric features—length (L) and width (W), (3) feature computation, (4) data partitioning for training and testing, (5) classification using Support Vector Machine (SVM), Random Forest (RF), and K-Nearest Neighbor (KNN), and (6) evaluation using a confusion matrix. Among all models tested, the Random Forest algorithm yielded the highest accuracy at 93%, with classification results showing True Positive = 349, False Positive = 28, True Negative = 223, and False Negative = 0. The findings highlight RF’s potential for enhancing automated fish monitoring in smart aquaculture systems.