海岸带地物的面向对象分类算法研究

发布时间：2018-11-07 16:36

【摘要】：遥感因其信息量大、获取信息速度快、周期短、受限制少等优势,为海岸带调查提供了一种新的手段。其中对海岸带遥感影像的解译是遥感于海岸带调查最常见的应用。目前,面向对象的分类方法已经成为遥感图像解译的新方向。该方法首先将影像分割成一系列彼此相邻的同质区域,即对象,之后将对象分类。与传统的基于像元的方法相比,对象蕴含更丰富的形状、纹理信息,能够更全面地描述图像特征,从而提高分类精度。近年来,有学者将面向对象的分类方法应用于海岸带地物的分类。然而,海岸带的地物类型复杂,自然与人工地物并存,它们的光谱、纹理、形状、分布特征各异。目前仍没有一种方法能够准确提取所有类型的地物。因此,本文以“海岸带地物的面向对象分类算法研究”为题,做了如下工作：1、针对海岸带影像的自身特点确定地物类别,将纹元理论引入海岸带的面向对象分类之中,提出一种基于纹元的面向对象分类；2、采用结构张量对基于纹元的面向对象分类做出优化。使用结构张量的迹提取梯度图像,优化图像分割,采用结构张量纹理构建纹元优化分类过程；3、局部二值模式(LBP)是一种简单有效的纹理特征提取算法,笔者将针对LBP在海岸带面向对象分类中的应用展开研究。基于LBP纹理生成标记图像,进行分水岭分割,挖掘对象的几何特征信息,最后使用BP神经网络进行分类。结果表明：纹元在利用图像光谱特征的同时挖掘了图像的纹理特征,可以很好地表现图像特征,基于纹元的面向对象分类是一种有效的分类方法；结构张量可以对基于纹元的分类模型从分割和分类做出优化,从而优化整个分类过程；LBP对于某些海岸带地物能够进行很好地区分,对海岸带地物的分类提取有很大的研究价值。
[Abstract]:Remote sensing provides a new method for coastal zone investigation because of its advantages of large amount of information, fast speed of obtaining information, short period and limited limit. The interpretation of coastal zone remote sensing image is the most common application of remote sensing in coastal zone survey. At present, object-oriented classification method has become a new direction of remote sensing image interpretation. The method first divides the image into a series of adjacent homogeneous regions, that is, objects, and then classifies the objects. Compared with traditional methods based on pixel, objects contain more shape and texture information, which can describe image features more comprehensively and improve classification accuracy. In recent years, some scholars have applied the object-oriented classification method to the classification of coastal features. However, the types of coastal features are complex, natural and artificial features coexist, their spectra, textures, shapes and distribution characteristics are different. At present, there is no method to extract all kinds of features accurately. Therefore, this paper focuses on the research of object-oriented classification algorithm for coastal features, and does the following work: 1. According to the characteristics of coastal zone images, the classification of ground objects is determined, and the grain element theory is introduced into the object-oriented classification of coastal zones. An object oriented classification based on grain element is proposed. 2. Structure Zhang Liang is used to optimize object-oriented classification based on grain element. The gradient image is extracted by using the trace of structure Zhang Liang, the image segmentation is optimized, and the texture of the texture is used to construct the texture element to optimize the classification process. 3. Local binary pattern (LBP) is a simple and effective texture feature extraction algorithm. The author will study the application of LBP in coastal zone object-oriented classification. Marking images are generated based on LBP texture, watershed segmentation is carried out, geometric feature information of objects is mined. Finally, BP neural network is used for classification. The results show that the texture feature of the image can be well represented by using the spectral feature of the image, and the object oriented classification based on the grain element is an effective classification method. Structure Zhang Liang can optimize the classification model based on grain element from segmentation and classification, thus optimizing the whole classification process. LBP can distinguish some coastal features well and has great research value for classification and extraction of coastal zone features.
【学位授予单位】：国家海洋环境预报中心
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：P715.7

【相似文献】