基于优化支持向量机的大豆品质检测研究  

Research on soybean quality detection based on optimized support vector machines

作　　者：姚立栋 秦华伟[1] 杨露露 YAO Lidong;QIN Huawei;YANG Lulu(School of Mechanical Engineering,Hangzhou Dianzi University,Hangzhou Zhejiang,310018,China)

机构地区：[1]杭州电子科技大学机械工程学院,浙江杭州310018

出　　处：《杭州电子科技大学学报(自然科学版)》2025年第1期67-73,共7页Journal of Hangzhou Dianzi University:Natural Sciences

摘　　要：为了提高计算机视觉对大豆品质检测的准确率和检测速度,提出基于麻雀搜索算法(Sparrow Search Algorithm, SSA)优化支持向量机(Support Vector Machine, SVM)及通过主成分分析法(Principal Component Analysis, PCA)进一步优化SSA-SVM模型的两种大豆品质检测模型。采集正常、破损和霉变的大豆图像并进行预处理,提取HSV颜色空间下的12个颜色特征参数,基于灰度共生矩阵法提取4个纹理特征参数。利用SSA寻找SVM的最佳惩罚因子和核函数半径。结果表明,SSA-SVM模型的准确率达到98.00%,较SVM模型的95.33%提高2.67%;使用PCA将16个特征参数降维到8个,构建PCA-SSA-SVM模型,该模型准确率为97.33%,较SVM模型的95.33%提高2.00%。PCA-SSA-SVM模型与SSA-SVM模型相比,其准确率低了0.67%,但迭代时间为11.78 s,较SSA-SVM的13.72 s缩短14.14%。结果表明,SSA-SVM模型与PCA-SSA-SVM模型在大豆品质检测方面均优于SVM模型。In order to improve the accuracy and speed of soybean quality detection by computer vision,two soybean quality detection models,based on Sparrow Search Algorithm(SSA)optimized Support Vector Machine(SVM)and Principal Component Analysis(PCA)further optimized SSA-SVM model are proposed.Normal,damaged and mildewed soybean images were collected and preprocessed to extract 12 color feature parameters in HSV color space,and 4 texture feature parameters were extracted based on gray co-existence matrix method.The SSA was used to find the optimal penalty factor and kernel radius of the SVM,The results show that the accuracy of SSA-SVM model is 98.00%,which is 2.67%higher than that of SVM model(95.33%).PCA was used to reduce the dimension of 16 characteristic parameters to 8,and the PCA-SSA-SVM model was constructed.The accuracy of the model was 97.33%,which was 2.00%higher than that of the SVM model(95.33%).Compared with the SSA-SVM model,the accuracy is 0.67%lower,but the iteration time is 11.78 s,which is 14.14%shorter than the 13.72 s before dimension reduction.The results show that both the SSA-SVM model and the PCA-SSA-SVM model are superior to the SVM model in soybean quality detection.

关 键 词：大豆 品质检测 计算机视觉 麻雀搜索算法 支持向量机 主成分分析法 

分 类 号：S565.1[农业科学—作物学]