基于机器视觉与鲍鱼足部纹理特征的公母分类研究OA

To address the current issues of low efficiency,high cost,and insufficient accuracy in manual sex classification of abalone,a classification model based on the DPO-SVM algorithm is proposed.By establishing a machine vision system,we collected foot texture images from 800 abalone specimens.Using the Gray-Level Co-occurrence Matrix(GLCM),we extracted a three-dimensional feature subset comprising energy(ASM),entropy(ENT),and contrast(CON)as optimal inputs.The classification accuracy on the test set reached 80.25％.To address the challenge of SVM hyperparameter optimization,particle swarm optimization(PSO)and whale optimization algorithm(WOA)were further integrated.This established a collaborative mechanism driven by the DPO algorithm for global search and local exploration,resolving the issues of local convergence or insufficient accuracy inherent in single algorithms.Results demonstrate that the DPO algorithm achieved an optimal fitness of 98.33％,surpassing PSO by 3.33 percentage points and WOA by 6.66 percentage points.Convergence was reached in just 6 iteration steps,reducing the training time cost by 57.14％compared to PSO.The DPO-SVM model achieved 100％overall classification accuracy,surpassing traditional SVM by 21.2％and eliminating selective misclassification of female abalone by single-population optimization algorithms.It maintained over 96％stability under complex conditions such as fluctuating lighting and minor noise.Research indicates that this algorithm combines high accuracy with low computational time,offering a reliable technical solution for the recognition module of future automated abalone sorting equipment in the aquaculture industry.It holds significant theoretical value for small-to-medium sample size aquatic classification scenarios.