Rapid Identification Method of Chinese Mitten Crab Based on Attention Mechanism and Deep Neural Network
We established a mathematical model for predicting the health status of Chinese mitten crab based on deep inference model (YO-LO-v7).Firstly,crabs grew in the natural environment form the back pattern characteristics,which could be divided into six features of lateral teeth,keel ridge,frontal gibbosity,verruca process,neck groove,compound eye according to morphometrics.Based on the human visual attention mechanism,the effective feature characterizations were visualized with higher classification accuracy in the YOLO-v7 model.Moreover,accord-ing to the calculation results,the image labeling software-LabelImg was used to mark the vitality grade of the first five different feature combina-tion modes,respectively.Then,the YOLO-v7 model was used to train and reason the marked data,and the optimal Chinese mitten crab freshness identification model was obtained.The experimental results showed that the proposed texture feature combination algorithm of verruca process+cervical groove could basically realize the recognition of the health status of Chinese mitten crab. The overall training accuracy could reach 95％,the reasoning accuracy could reach 96.20％.Moreover,the reasoning time of each vitality grade of Chinese mitten crab was less than one second.This method had great application prospect and market value,which provided key technology for developing nondestructive testing e-quipment for large-scale online quality of Chinese mitten crab.
Chinese mitten crabRapid identificationHealth statusYOLO-v7Combination feature
万方数据
维普
