基于人工免疫算法的储粮害虫特征选择研究

储粮害虫特征选择是粮虫图像识别中一个关键的预处理环节.提出基于v折交叉验证训练模型识别率和所选特征个数的特征子集评价准则,将人工免疫算法应用到粮虫的特征选择.该算法从粮虫的17雏形态学特征中自动选择出面积、周长等7个特征的最优特征子空间,采用参数优化之后的SVM分类器对90个粮虫样本进行分类,识别率达到95.5%以上,并与PCA法、GA法和原始特征法进行了对比,结果表明人工免疫算法降低了特征空间的维数,提高了分类器的识别率,证实了基于人工免疫算法的粮虫特征选择是可行的.

Feature Selection for Stored-Grain Insects Based on Artificial Immune Algorithm

he feature selection is a key pre-processing point in the image recognition of stored-grain insects.The performance evaluator of the feature selection was proposed based on the recognition accuracy of the v-fold cross-validation training model and the number of the feature subset.The artificial immune algorithm was applied to the feature selection of stored-grain insects.The algorithm selected 7 features that comprised the optimal feature space,such as area and perimeter,from 17 morphological features.The support vector machine classifier with optimized parameters automatically recognized ninety image samples of stored-grain insects from a grain-depot,and the correct identification rate was over 95.5%.The artificial immune algorithm was compared with the methods of principal component analysis,genetic algorithm and original feature,Results:The artificial immune algorithm reduces the feature dimensions and improves the recognition accuracy.It is proved that the artificial immune algorithm is practical and feasible.