项目名称: 面向图像分割的自适应脉冲耦合神经网络理论及应用研究
项目编号: No.61201363
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 电子学与信息系统
项目作者: 姚畅
作者单位: 北京交通大学
项目金额: 24万元
中文摘要: 脉冲耦合神经网络来源于高级哺乳动物的视觉仿生,更符合人类的视觉感知机理,近年来受到了广泛关注,为提高图像处理性能带来了新的机遇。尽管现有理论取得了初步成果,但由于网络参数自动设置等难题,限制了其进一步的应用和发展。 本项目以建立面向图像分割的自适应脉冲耦合神经网络模型为目的,深入研究脉冲耦合神经网络神经元的优化模型,网络参数的自动确定和寻优算法,控制图像分割质量的迭代点火终止准则,并在此基础上构建适用于图像分割的自适应脉冲耦合神经网络模型,提高图像分割的精度和鲁棒性,为基于人类视觉特性的图像分割方法的研究提供新的解决思路和解决方法。本项目的研究将对图像处理、人工智能、计算机视觉等领域的发展具有重要意义,对人工神经网络及其相关学科的发展也将起到积极的推动作用。
中文关键词: 图像分割;脉冲耦合神经网络;参数确定;视觉注意机制;神经元
英文摘要: Pulse coupled neural network (PCNN) is a bio-vision bionics from senior mammalian with more similarity to the human visual perception mechanism, which has been accepted great attentions in recent years, and brings new opportunities for improving image processing performance. Although existing PCNN theory made preliminary results, there are still many problems, such as parameters automatically set, blocked the PCNN's deeper application and development. In order to model adaptive PCNN for image segmentation, this project will explore optimal PCNN neuron model, the rule of network parameters automatic determination, optimizing algorithms, terminate conditions for controlling the quality of segmented image and propose the adaptive PCNN model which could improve the precise and robust for image segmentation. The research of project will provide a new solution and method for the study of image segmentation based on human visual perception, and will be significant to image processing, artificial intelligence, computer vision domains' development, and also will motivate the development of artificial neural network discipline.
英文关键词: Image Segmentation;Pulse Coupled Neural Network;Parameter Determination;Visual Attention Mechanism;Neuron