基于视觉感知原理的高分辨率遥感影像分割与人工目标提取研究

发布时间：2018-12-17 05:30

【摘要】：随着传感器技术的不断发展,近些年来,遥感影像所获取的地面分辨率不断提高。因此,越来越多的地面细节被遥感影像清晰的表现出来,遥感影像对于地面细节的描述能力不断提高。这些高分辨率遥感影像由于自身具有较高的空间分辨率以及较少的光谱分辨率,导致传统的遥感影像处理技术无法对其进行较好的分析与处理,对遥感影像的处理技术提出了新的挑战。然而,对于空间分辨率很高的遥感影像,通过人眼的观察,可以被轻松地辨识与认知影像中多种多样的地物及其复杂的细节。人类的视觉感受能够从颜色、纹理、和形状等方面对高分辨率遥感影像进行全方位的观察与认知。根据这一特点,本文试图通过模拟人眼视觉感知的图像处理能力,来对高分辨率遥感影像进行认知与识别。具体来说,主要包含以下几个方面：1 以视觉神经感受理论为基础,提出了模拟视觉感受的高分辨率遥感影像分割方法。该方法从多个视觉感知特点出发,通过无监督的数学和影像处理工具,结合实际高分率遥感影像的表现,对遥感影像进行光谱、纹理、细节的详细分析,建立模拟人眼视觉感知能力的遥感影像分割模型,获取区域一致性较好而且细节刻画能力强的分割结果,从而实现对高分辨率遥感影像的认知与分析。并根据实际试验来对分割模型中的多个参数取值进行测试、分析与归纳,得到参数取值的规律。以真实的高分辨率卫星遥感影像进行了分割实验,相较于传统算法,本文提出的方法具有最好的分割能力,分割结果与人眼目视感觉相似,证明了本文提出的方法切实有效。2 以道路特用的形状结构为基础,提出了基于视觉感知的形状分析道路提取方法。该方法的根据高分率遥感影像上的道路特点,从形状的角度出发,对影像上疑似道路的地面目标进行分析,从中找到具有长直线性以及网状线性结构的道路目标,并根据视觉感知心理学格式塔原理,建立了形状分离模型,从而对非线性的非道路信息进行分离,最终得到完整的道路信息。并根据实际试验来对道路提取方法中的不同参数取值进行测试、分析与归纳,得到参数取值的规律。在道路分布复杂的城区高分辨率航空影像上进行了道路提取实验,对比参考道路信息,本文方法能够对道路进行正确认知,证明了本文提出的道路提取算法能够从复杂地面环境中快速、可靠、稳定地提取道路。3根据人眼对形状的基础认知,提出了基于视觉感知的建筑物形状分析方法。该方法从人眼对于形状认知感受出发,结合建筑物在遥感影像上的面状形状表现,从多个形状角度以多个参数的方式,分别独立地对建筑物形状进行刻画与认知,并根据多个参数的不同表现,对建筑物进行形状认知。并通过一系列参数性能测试,对不同的参数进行了形状认知能力的测试、评估与分析,对各参数的特点与使用方法进行了归纳和总结,并根据不同参数的性能,制定了科学合理的形状认知方法。在使用真实城区高分辨率航空影像进行了基于形状的建筑物提取实验,通过对影像的分割结果进行形状分析,可以获得与人眼视觉感受一致的建筑物提取结果,证明了本文提出的建筑物形状认知参数能够简单、准确地对建筑物地物进行刻画与认知。
[Abstract]:With the development of sensor technology, the ground resolution acquired by remote sensing image has been improved in recent years. As a result, more and more ground detail is clearly displayed by the remote sensing image, and the ability of the remote sensing image to describe the details of the ground is increasing. The high-resolution remote sensing image has high spatial resolution and less spectral resolution, leading to the fact that the traditional remote sensing image processing technology can not perform better analysis and processing on the high-resolution remote sensing image, and a new challenge is put forward for the processing technology of the remote sensing image. However, for remote sensing images with high spatial resolution, through the observation of the human eye, it is possible to easily recognize and recognize a wide variety of features and their complex details in the cognitive image. The human visual sense can make all-round observation and cognition of the high-resolution remote sensing image from the aspects of color, texture, and shape. According to this feature, this paper attempts to recognize and recognize the high-resolution remote sensing image by simulating the image processing capability of the human eye. In particular, it mainly includes the following aspects: 1. Based on the visual nerve sense theory, a high-resolution remote sensing image segmentation method for simulating the visual perception is presented. The method comprises the following steps of: starting from a plurality of visual perception characteristics, carrying out detailed analysis on the spectrum, the texture and the detail of the remote sensing image through the non-supervised mathematical and image processing tools, and establishing a remote sensing image segmentation model for simulating the visual perception capability of the human eyes, and obtaining the segmentation result with good regional consistency and strong detail description ability, so as to realize the recognition and analysis of the high-resolution remote sensing image. and according to the actual test, the values of a plurality of parameters in the segmentation model are tested, analyzed and summed to obtain the law of the value of the parameters. based on the shape and structure of the road, the method has the best segmentation ability and the segmentation result is similar to that of the human eye. A method of shape analysis for road extraction based on visual perception is proposed. according to the characteristic of the road on the high-score remote sensing image, the ground target of the suspected road on the image is analyzed from the angle of the shape, the road target with the long-line property and the mesh-like linear structure is found, and the shape separation model is established, so that the non-linear non-road information is separated, and the complete road information is finally obtained. and according to the actual test, the values of different parameters in the road extraction method are tested, analyzed and summed to obtain the law of the value of the parameters. The road extraction experiment is carried out on the high-resolution aerial image of a complex urban area with complex road distribution, and the road information is compared with the reference road information. The method can correctly recognize the road, and proves that the road extraction algorithm proposed by the invention can be fast and reliable in a complex ground environment, The method of building shape based on the visual perception is put forward based on the human eye's basic cognition of the shape. According to the method, the shape of a building is characterized and recognized in a plurality of parameters from a plurality of shape angles in a manner of a plurality of parameters from a plurality of shape angles, and the shape of the building is recognized. Through a series of parameter performance tests, the test, evaluation and analysis of the shape cognition ability of different parameters are carried out, the characteristics and the methods of using the parameters are summarized and summarized, and a scientific and reasonable shape cognition method is developed according to the performance of different parameters. in that invention, a shape-based building extraction experiment is carry out on the high-resolution aerial image of the real urban area, the shape analysis is carried out on the segmentation result of the image, and a building extraction result which is consistent with the visual perception of the human eye can be obtained, It is proved that the building shape cognitive parameters proposed in this paper can be used to describe and cognize the building features in a simple and accurate way.
【学位授予单位】：武汉大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：P237

【相似文献】