中国科学院光电技术研究所机构知识库
Advanced  
IOE OpenIR  > 光电技术研究所博硕士论文  > 学位论文
题名:
基于分形理论的目标检测分割技术
作者: 宿丁
学位类别: 博士
答辩日期: 2007-05-31
授予单位: 中国科学院光电技术研究所
授予地点: 光电技术研究所
导师: 张启衡
关键词: 分形 ; 目标检测 ; 目标分割 ; 模式识别 ; 图像处理
其他题名: Object Detection and Segmentation Technology
学位专业: 信号与信息处理
中文摘要: 在数字图像处理中引入分形理论,使传统的基于人造目标的光电探测技术转向基于自然背景的研究,开拓了图像处理的新思路, 丰富了目标检测的手段。本论文旨在研究和发展分形理论在图像检测分割过程中的应用技术及相关技术,为多种复杂环境条件下多种类型目标的检测难题提供解决方法和技术途径。 本论文紧密围绕分形处理这一核心,在系统地阐述分形的基本理论和方法的基础上,从理论和实现两个方面对该领域所面临的主要问题进行了深入探讨,提出了若干新的分形应用算法和技术,在检测精度和适应能力等方面取得了较大进展。 在分形适用性方面,系统研究不同背景、目标和噪声满足分形集合的整体趋势,分析了线性无标度区间对分形参数的影响。定义分形模型吻合度特征,建立了具有同类分形特性的背景之间的联系,不仅为衡量自然背景的复杂程度提出了一种客观的评判标准,也为应用分形技术构建了总体处理框架。 在预处理技术方面,详细分析灰度梯度变化对图像分形特性的影响,提出了分形特征的增强策略。在控制目标信息丢失的前提下,既有效抑制了背景中因灰度起伏而产生的大量分形维数奇异区域,又凸显了目标和背景的空间结构特性差异,为复杂背景下目标的分形检测开辟了一条可行的预处理途径。 分形检测分割新技术是论文的研究重点。针对不同类型背景建立了相应的分形模型,结合目标特性设计了多种实用算法,完善了分形处理的流程体系。 (1)针对天空、海天等远距离低对比度场景,利用其较强的空域统计自相似性信息,通过动态规划思想改进DBC算法,提取分形模型拟合误差进行检测,再通过目标的聚类特征进行判别,达到了较高的检测概率和较低的虚警概率。 (2)针对地空等复杂场景,依据扩展目标边缘特性和串行分割模型提出了两种分割新算法。其一,根据目标和背景交错处边缘响应的幅值相似但方向有较大差异这一特性,结合多尺度滤波和频域极大似然估计的方法,在大尺度下提取目标的主要轮廓,根据边缘方向信息,选择与小尺度下提取的幅度和方向接近的目标边缘进行连接,在复杂自然条件下提取完整且闭合的扩展目标轮廓,从而提高了分割效率和精度。其二,引入极坐标系中优化的形态学分形模型,从集合角度细致分析了极坐标系中可变尺度形态结构算子的旋转缩放特性,在提取目标分形特征的同时实现对背景杂散纹理的滤波处理,再对分形特征进行边界跟踪,并综合复杂度先验信息抑制背景离散团块,保留并填充面积最大的目标团块,达到精确分割的目的。 (3)针对单传感源中海杂波场景下的多目标,首先利用尺度误差检测并标记有效视场内的目标个数及所在区域,再根据熵相似准则进行区域生长,在潜在目标区域中提取出目标的精确位置与完整轮廓;针对多传感源中复杂场景下的多目标,首先通过最大熵阈值分割出红外图像的潜在目标区域,记录其质心位置及形状大小并对应到可见光图像中,再提取可见光图像分形维数的方差特征,对记录的目标区域进行初判决,得到真实目标质心处的分形维数方差,然后将分形维数图划分为与已知目标大小接近的区域块,搜索并标记具有相近方差特征的所有区域块,最后在已标记及其相邻的区域块中精确分割出全部目标。 以上算法在一系列仿真实验中均取得良好的分割效果。本论文采用客观评价函数评估了分形算法的检测概率、错(漏)检概率和线性连接度等,而且通过数据学习机制,分层逐级缩小候选目标区域,降低了算法的计算量。 作为预研技术课题,本论文还探索了一条有效的模型验证、算法评估的开发途径,为图像处理系统设计、验证和实现提供了新的手段。通过优化分形特征和计算步骤,利用Matlab S-function封装Simulink模块,搭建了可现场采集、处理、测试的数字图像分形实时仿真平台。同时,本论文对未来的硬件平台进行了资源分析和规划;研究归整化对数查找表技术,解决了分形处理需要大量在线对数运算的难题;探索利用第三方工具直接将Simulink模型转化为硬件描述语言的可行性,为分形理论未来的广泛应用指明了方向。
英文摘要: The Ph.D article introduces fractal theory into the region of digital image processing. It explores new idea of image processing and makes classical electro-optical detection technique based on man-made object turn to novel study based on nature background. The research is a pre-research item and aims at developing some applicable technologies and relative productions in object detection and segmentation. The thesis tightly focuses on fractal processing, explains basic fractal theory and methods, completely and deeply discusses main problems in practice, designes and progresses several novel and robust detection and segmentation algorithms and acquires evident improvements in adapt ability and detection precision. In the fact of applicability of fractal algorithm, the article analyses the trend of different background, object and noise fitting for fractal aggregate, studies the question of non-scale interval on the mathematical model, firstly defines a new concept of Fractal Fitting Degree, which not only could weigh complexity of background in an impersonality criterion, but also establish the relationship between similar fractal characters and construct main frame of fractal processing. In the fact of image pre-processing, the article particularly analyses the effect of gray intensity variety, advances ideas to improve characteristics. It restrains oddity fractal dimension values and protrudes diversities of scene and object, pioneers an easy and accessible way to detect objects hiding in compound background. The emphasis of the research is novel detection and segmentation technologies. To deal with long distance and low contrast backgrounds such as sky or sea, using the obvious self-similar stationary information of them, mends Differential Box-Counting algorithm, computes out fractal fitting errors to detect and distinguish real objects. This way can achieve higher detection probability and lower mistake probability. To deal with complex ground scenes, two fresh algorithms are proposed from contour characteristics. The first one is based on the specialty of different edges which has similar amplitude but diversity direction, which combines with multi-scale filter and maximum likelihood estimation in frequency region, computes out main object contour in large scale, chooses edges with nearly same amplitude and direction in small scale to link together and obtains full and closed figure. The other one inherits fractal morphology model in polar coordinate, analyses multi-scale morphology operator, abstracts fractal feature and restrains small texture of background and at last reserves and fills in the block with the largest area to get the whole object. To deal with multi-object from single detector, it uses feature named scale error to mark the number and position of object, according to entropy rules to acquire all the objects. Besides that, a novel scheme is proposed to deal with multi-objects in complex background from multi-detector. At first, it segments region which may contain real objects in infrared image based on maximum entropy rule, register centroid positions and finds them out in visible light image. Second, abstracts stationary fractal feature to judge centroids of real objects, divides the light image into blocks as the same size of true objects. At last, it searches all of the objects and segments out. These algorithms acquire good effect in a serial of emulation experiments. The article explains their abilities by designing impersonal estimation function. However, many ideas are used to decrease compound computation. As a pre-research projects, the paper explores an efficient way from system design to system implement, helps researchers to pay more attention to algorithm design but not to implement ways. It utilizes Matlab S-function to encapsulate Simulink modules, constructs a real-time emulation and implement platform. In the same time, it analyses resources of hardware platform and solves the question of logarithm operation, finds ways to transform Simulink models directly into VHDL or Verilog language, demonstrates the direction of further wide study and application of fractal technology.
语种: 中文
内容类型: 学位论文
URI标识: http://ir.ioe.ac.cn/handle/181551/230
Appears in Collections:光电技术研究所博硕士论文_学位论文

Files in This Item:
File Name/ File Size Content Type Version Access License
10001_200418015127026宿丁_paper.doc(10399KB)----限制开放View 联系获取全文

Recommended Citation:
宿丁. 基于分形理论的目标检测分割技术[D]. 光电技术研究所. 中国科学院光电技术研究所. 2007.
Service
Recommend this item
Sava as my favorate item
Show this item's statistics
Export Endnote File
Google Scholar
Similar articles in Google Scholar
[宿丁]'s Articles
CSDL cross search
Similar articles in CSDL Cross Search
[宿丁]‘s Articles
Related Copyright Policies
Null
Social Bookmarking
Add to CiteULike Add to Connotea Add to Del.icio.us Add to Digg Add to Reddit
文件名: 10001_200418015127026宿丁_paper.doc
格式: Microsoft Word
此文件暂不支持浏览
所有评论 (0)
暂无评论
 
评注功能仅针对注册用户开放,请您登录
您对该条目有什么异议,请填写以下表单,管理员会尽快联系您。
内 容:
Email:  *
单位:
验证码:   刷新
您在IR的使用过程中有什么好的想法或者建议可以反馈给我们。
标 题:
 *
内 容:
Email:  *
验证码:   刷新

Items in IR are protected by copyright, with all rights reserved, unless otherwise indicated.

 

 

Valid XHTML 1.0!
Copyright © 2007-2016  中国科学院光电技术研究所 - Feedback
Powered by CSpace