南京理工大学学报（自然科学版）
南京理工大學學報（自然科學版）
남경리공대학학보（자연과학판）
JOURNAL OF NANJING UNIVERSITY OF SCIENCE AND TECHNOLOGY
2014年
4期
558-565
,共8页
脉冲耦合神经网络%人工蜂群算法%人工蜂群算法-脉冲耦合神经网络模型%乘积型交叉熵%图像分割
脈遲耦閤神經網絡%人工蜂群算法%人工蜂群算法-脈遲耦閤神經網絡模型%乘積型交扠熵%圖像分割
맥충우합신경망락%인공봉군산법%인공봉군산법-맥충우합신경망락모형%승적형교차적%도상분할
pulse coupled neural network%artificial bee colony%artificial bee colony-pulse coupled neural network%product cross entropy%image segmentation
为使标准脉冲耦合神经网络( Pulse coupled neural network,PCNN)模型在图像分割中能够自适应地调整模型参数与全局阈值,提高分割效果,该文提出一种基于人工蜂群( Artificial bee colony,ABC)算法改进的自适应PCNN 模型,即人工蜂群算法-脉冲耦合神经网络( ABC-PCNN)模型；提出了改进后的乘积型交叉熵函数,并利用ABC算法将此函数作为其适应度函数优化输出其连接系数和阈值。采用Lena图像和血细胞图像评估PCNN模型和ABC-PCNN模型的性能。实验结果表明：ABC-PCNN模型对图像的自适应分割效果优于PCNN模型。针对血细胞分割图像中存在的重叠区域,该文结合角点和质点坐标定位重叠区域的二次分割线得到最终分割图像,所提算法高效且能得到较好的分割结果。
為使標準脈遲耦閤神經網絡( Pulse coupled neural network,PCNN)模型在圖像分割中能夠自適應地調整模型參數與全跼閾值,提高分割效果,該文提齣一種基于人工蜂群( Artificial bee colony,ABC)算法改進的自適應PCNN 模型,即人工蜂群算法-脈遲耦閤神經網絡( ABC-PCNN)模型；提齣瞭改進後的乘積型交扠熵函數,併利用ABC算法將此函數作為其適應度函數優化輸齣其連接繫數和閾值。採用Lena圖像和血細胞圖像評估PCNN模型和ABC-PCNN模型的性能。實驗結果錶明：ABC-PCNN模型對圖像的自適應分割效果優于PCNN模型。針對血細胞分割圖像中存在的重疊區域,該文結閤角點和質點坐標定位重疊區域的二次分割線得到最終分割圖像,所提算法高效且能得到較好的分割結果。
위사표준맥충우합신경망락( Pulse coupled neural network,PCNN)모형재도상분할중능구자괄응지조정모형삼수여전국역치,제고분할효과,해문제출일충기우인공봉군( Artificial bee colony,ABC)산법개진적자괄응PCNN 모형,즉인공봉군산법-맥충우합신경망락( ABC-PCNN)모형；제출료개진후적승적형교차적함수,병이용ABC산법장차함수작위기괄응도함수우화수출기련접계수화역치。채용Lena도상화혈세포도상평고PCNN모형화ABC-PCNN모형적성능。실험결과표명：ABC-PCNN모형대도상적자괄응분할효과우우PCNN모형。침대혈세포분할도상중존재적중첩구역,해문결합각점화질점좌표정위중첩구역적이차분할선득도최종분할도상,소제산법고효차능득도교호적분할결과。
In order to adjust the model parameters and the global threshold for image segmentation, an improved pulse coupled neural network ( PCNN ) model based on artificial bee colony ( ABC ) algorithm,namely ABC-PCNN,is proposed here. It combines a new criterion of product cross entropy with the standard simplified PCNN model. The product cross entropy is used as the fitness function to optimize the connection output coefficient and threshold value by the ABC algorithm. Lena image and blood cell image are used to evaluate the PCNN model and the ABC-PCNN model respectively. The experimental results show that the adaptive image segmentation by the ABC-PCNN model outperforms that by the PCNN model. As the overlapping areas need secondary segmentation in the segmented blood cell image,corners and center coordinates are used to locate the dividing line and to get the final image segmentation. The method proposed here is effective and can obtain better segmentation results.
