基于最小交叉熵的改进PCNN杨梅图像分割算法  

Improved PCNN bayberry image segmentation algorithm based on minimum cross entropy

作　　者：徐黎明[1,2] 吕继东[1] 

机构地区：[1]常州大学信息科学与工程学院,江苏常州213164 [2]江苏城乡建设职业学院建筑设备系,江苏常州213147

出　　处：《西北师范大学学报（自然科学版）》2016年第1期43-46,52,共5页Journal of Northwest Normal University(Natural Science)

基　　金：江苏省自然科学青年基金资助项目(BK20140266);江苏省高校自然科学研究面上项目(14KJB210001);江苏省高等职业院校国内高级访问学者计划资助项目(2014FX031);常州大学科研启动费资助项目(ZMF13020019)

摘　　要：针对传统脉冲耦合神经网络(Pulse coupled neural network,PCNN)模型复杂、迭代次数需要人工试验调整等缺点,对PCNN模型进行改进,提出了一种基于色差分量R-G和改进PCNN模型的杨梅图像分割方法.首先用杨梅图像的R-G色差分量作为输入;然后选用最小交叉熵准则,自适应生成PCNN的迭代次数;最后通过PCNN迭代实现杨梅图像的自动分割.实验结果表明,此方法与传统方法相比,能够有效避免背景干扰,提升了图像分割质量,分割结果更符合杨梅果实的特征.For the shortcomings of traditional pulse coupled neural net,such as model complexity,the number of iterations need to be manually adjusted,in this paper,PCNN model is improved,and a segmentation method based on R-G color component and improved PCNN model is put forward.Firstly,R-G color component of the bayberry image is used as input based on a simplified PCNN model.Then,the minimum cross entropy criterion is used to generate the iterations of PCNN.Finally,the bayberry image segmentation is achieved automatic with PCNN iteration.The results show that this method could effectively avoid background noise and enhance the quality of image segmentation compared with the traditional PCNN model algorithms,and make the results of the segmentation more in line with the characteristics of bayberry fruit.

关 键 词：脉冲耦合神经网络 颜色特征 最小交叉熵 图像分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]