A Novel and Non-Invasive Approach to Estimate Abalone Biomass Using Image Analysis

Physical removal of abalone from the attached substratum for biomass assessment can cause stress, physical damage or even death. This research aims to develop and test 3 models using an image-based method to accurately determine weight and length for 3 age classes of cultured abalone. Abalone were grouped into three size classes (1, 2, and 3 years old) for photography. Photos were analyzed using image processing. The extracted data were used to develop the prediction models and perform a validation test using an independent dataset. In the 1-year-old group, the quadratic model was a fitted regression for estimating the weight of abalone based on the image of the shell, using shell length as the predictor. In contrast, the power model was able to accurately predict the weight using the shell area. In the 2-year-old group, the lowest estimation error values were observed using the shell area quadratic model. The quadratic model was better at predicting abalone weight from either image-based shell length or shell area in the 3-year-old group. The power regression model exhibits better performance in weight estimation, regardless of abalone size, using the image-based shell length. Either the quadratic model or the power model can be used to estimate the weight of farmed abalone using an image-based shell area. Our results indicate that the success of image analysis technology to estimate abalone mass could reduce negative impacts on abalone during biomass determination and contribute to the efficiency of management in abalone farming. The models could potentially be applied in stocking assessment work with some modifications.