Aquaculture fish counting and mass estimation method via vibration signal processing

Abstract

Information on the mass and counts of fish at different growth stages is crucial for precision feeding management and the improvement of fish farming environments. Grading operation plays a significant role in management and harvesting of cage culture. In this study, a counting and mass estimation of fish via vibration signal process method was proposed, using a roller-type fish grader to count and estimate the mass of fish during the grading process. First, we replicate fish dropping events during the grading process, when fish impact the mass measurement plate, on a test bench, and waveform data from vibration sensors located on the mass measurement plate were collected. Secondly, the waveform signal was analyzed, and trigger thresholds were set. When the signal exceeded the threshold, performed mathematical operations on the signal's data for counting and mass estimation until the signal left the threshold. The method was conducted on a test bench, and the results showed a coefficient of determination (R²) of 0.9970 and a Root Mean Square Error (RMSE) of 4.1724 g for mass estimation. Furthermore, the method was validated on a roller-type fish grader in practical applications, achieving average R² of 0.9862 and RMSE of 8.8529 g. The accuracy of fish counting reached 100 %. These experimental results demonstrate the proposed method can accurately count and estimate mass of fish.