Research and Application of Key Technologies for Crayfish Identification and Measurement
The phenotypic traits of crayfish provide an important economic basis in aquaculture and genetic breeding.In order to accurately realize the visual detection and positioning of crayfish and measure the phenotypic characteristics of its main body,this paper first proposed a method for detecting crayfish based on the Mask RCNN instance segmentation model,which could not only quickly identify crayfish but also perform instance segmentation on it,so as to generate high-quality binary mask maps.On this basis,a new crayfish measurement algorithm was proposed.Specifically,by extracting the contour curve of crayfish,the central axis was obtained,and the partition method was used to extract the characteristic area of crayfish.Then,the convex hull and convex hull defect algorithms were utilized to obtain the feature points of the crayfish,thus realizing the contactless measurement calculation of the crayfish's main parts such as its total length,body length,carapace,abdomen,and tail fan.The results show that the Mask RCNN model has good generalization performance on the crayfish dataset,and the accuracy of model segmentation reaches 94.6%.The average detection accuracy of target recognition reaches 98.7%,and the average absolute error of each measurement index is within 5 mm.The Mask RCNN model is easier than manual measurement,and it is stable and efficient,with better repeatability.This method is beneficial to the identification and rapid acquisition of the structure size of crayfish in the process of production and breeding.
crayfishMask RCNNconvex hull and convex hull defectcentral axiscontactless measurement calculation
NETL
NSTL
万方数据
