计算机工程与应用
計算機工程與應用
계산궤공정여응용
COMPUTER ENGINEERING AND APPLICATIONS
2014年
16期
188-191
,共4页
图像识别%特征选择%小波矩%入侵性杂草优化算法
圖像識彆%特徵選擇%小波矩%入侵性雜草優化算法
도상식별%특정선택%소파구%입침성잡초우화산법
image recognition%features selection%wavelet moment%invasive weed optimization algorithm
针对小波不变矩提取的特征向量维数过大的问题,提出一种以类间、类内散布矩阵作为可分离判据的离散入侵性杂草优化算法实现特征向量的选择,利用BP神经网络作为分类器进行图像识别。实验仿真结果表明,与现有特征选择算法相比,改进的离散入侵性杂草优化算法对于图像特征向量的选择时间更短,识别正确率更高,能有效提高分类器的性能。
針對小波不變矩提取的特徵嚮量維數過大的問題,提齣一種以類間、類內散佈矩陣作為可分離判據的離散入侵性雜草優化算法實現特徵嚮量的選擇,利用BP神經網絡作為分類器進行圖像識彆。實驗倣真結果錶明,與現有特徵選擇算法相比,改進的離散入侵性雜草優化算法對于圖像特徵嚮量的選擇時間更短,識彆正確率更高,能有效提高分類器的性能。
침대소파불변구제취적특정향량유수과대적문제,제출일충이류간、류내산포구진작위가분리판거적리산입침성잡초우화산법실현특정향량적선택,이용BP신경망락작위분류기진행도상식별。실험방진결과표명,여현유특정선택산법상비,개진적리산입침성잡초우화산법대우도상특정향량적선택시간경단,식별정학솔경고,능유효제고분류기적성능。
Due to the large number of feature values which can be extracted from wavelet moment, a discrete invasive weed optimization algorithm is proposed to select the feature vectors with between-class scatter matrix and within-class scatter matrix, and finally can recognize images with the help of BP neural network as the classifier. The simulation results show that, compared to the feature selection algorithm, the improved discrete invasive weed optimization algo-rithm has the shorter selection time of image feature vectors, the higher accuracy, and can effectively improve the perfor-mance of the classifier.