基于AlexNet模型的大闸蟹自动分级系统设计与实现

Design and Implementation of the Chinese Mitten Crab Automatic Grading System Based on AlexNet Model

针对目前大闸蟹人工分级方法的局限性,设计基于Matlab图像处理的大闸蟹分级系统。首先,在湖州市太湖养殖基地采集不同等级大闸蟹背部和腹部图像,对采集的图像进行灰度化、阈值分割、形态学等预处理。然后利用卷积神经网络AlexNet模型提取大闸蟹公母特征,利用面积法计算其大小。通过选取的10只大闸蟹的重量和系统计算得到的像素转化为面积参数,分析得到大闸蟹背部图像像素占比与其重量成近似正比例关系,因此可根据背部图像的计算值得到其大小特征。根据大闸蟹公母、大小特征完成分级。实验结果表明,系统在大闸蟹公母识别方面平均准确率达到92.655%,大小分级方面平均准确率达到95%。

A Chinese mitten crab(Eriocheir sinensis)grading system based on Matlab image processing was designed to address the limitations of current manual grading methods for Chinese mitten crabs.First,the back and abdomen images of Chinese mitten crabs of different grades were collected at the the Taihu Lake breeding base in Huzhou City,and the collected images were preprocessed by graying,threshold segmentation,and morphology.Then,the convolutional neural network AlexNet model was used to extract the male and female features of Chinese mitten crabs,and its size was calculated using the Area Method.By selecting the weight of 10 Chinese mitten crabs and converting the pixels calculated by the system into area parameters,it was analyzed that the proportion of pixels in the back image of Chinese mitten crabs is approximately proportional to their weight.Therefore,their size characteristics can be obtained based on the calculated values of the back image.Grading was completed based on the male and female characteristics and size of Chinese mitten crabs.The experimental results show that the system has an average accuracy rate of 92.655% in recognizing male and female Chinese mitten crabs,with an average accuracy rate of 95% in size grading.