Accuracy Improvement of the Shrimp-size Estimation Method for an Automatic Feeding-tray Lifting System Used in Shrimp Farming

Abstract

Two shrimp-size estimation methods were previously proposed in the shrimp-growth estimation based on ResNeXt for an automatic feeding-tray lifting system. The first one is based on the size of the smallest rectangle in which a shrimp is inscribed. The second is based on the shrimp’s skeleton. Both methods have their strength and weakness. The former works better in the case of straight-body shrimp, but the latter works well in the case of curved-body shrimp. This study proposes a combination method of box-based and skeleton-based measurements by using a box ratio or eigenvalue ratio as a method-selection parameter to improve the performance of the shrimp-size estimation. Experimental results showed that the combination method using the box ratio can reduce the error by 63.56% and 9.24% compared with those of the previous box-based and skeleton-based methods, respectively. The combination method with the eigenvalue ratio can reduce the error by 63.10% and 8.08%, respectively. This study shows that combining the box-based and the skeleton-based measurements can reduce the error of the shrimp-size estimation method and improve its performance considerably.