• 欢迎访问速搜资源吧,如果在网站上找不到你需要的资源,可以在留言板上留言,管理员会尽量满足你!

【速搜问答】影像变化检测是什么

问答 admin 2周前 (04-19) 19次浏览 已收录 0个评论

汉英对照:
Chinese-English Translation:

遥感影像变化检测指利用多时相获取的覆盖同一地表区域的遥感影像及其它辅助数据来确定和分析地表变化。它利用计算机图像处理系统,对不同时段目标或现象状态的变化进行识别、分析;它能确定一定时间间隔内地物或现象的变化,并提供地物的空间分布及其变化的定性与定量信息。

Remote sensing image change detection refers to the use of multi temporal remote sensing images covering the same surface area and other auxiliary data to determine and analyze the surface change. It uses computer image processing system to recognize and analyze the changes of objects or phenomena in different periods of time; it can determine the changes of objects or phenomena in a certain time interval, and provide qualitative and quantitative information of spatial distribution and changes of objects.

遥感影像变化检测指利用多时相获取的覆盖同一地表区域的遥感影像及其它辅助数据来确定和分析地表变化。它利用计算机图像处理系统,对不同时段目标或现象状态的变化进行识别、分析;它能确定一定时间间隔内地物或现象的变化,并提供地物的空间分布及其变化的定性与定量信息。它的方法主要分为两大类:一类是基于像素的影像变化检测,一类是基于特征的影像变化检测。

Remote sensing image change detection refers to the use of multi temporal remote sensing images covering the same surface area and other auxiliary data to determine and analyze the surface change. It uses computer image processing system to recognize and analyze the changes of objects or phenomena in different periods of time; it can determine the changes of objects or phenomena in a certain time interval, and provide qualitative and quantitative information of spatial distribution and changes of objects. Its methods are mainly divided into two categories: one is pixel based image change detection, the other is feature-based image change detection.

定义

definition

遥感影像变化检测指利用多时相获取的覆盖同一地表区域的遥感影像及其它辅助数据来确定和分析地表变化。它利用计算机图像处理系统,对不同时段目标或现象状态的变化进行识别、分析;它能确定一定时间间隔内地物或现象的变化,并提供地物的空间分布及其变化的定性与定量信息。

Remote sensing image change detection refers to the use of multi temporal remote sensing images covering the same surface area and other auxiliary data to determine and analyze the surface change. It uses computer image processing system to recognize and analyze the changes of objects or phenomena in different periods of time; it can determine the changes of objects or phenomena in a certain time interval, and provide qualitative and quantitative information of spatial distribution and changes of objects.

通过不同时相遥感影像之间的分析,发现并提取它们之间的变化信息,是遥感变化检测的关键。影像变化检测(又指不同时相遥感影像之间的变化检测)的方法主要分为两大类:一类是基于像素的影像变化检测,一类是基于特征的影像变化检测。基于像素的影像变化检测,是在影像精确几何配准的基础上,对每个像素前后两个时相的灰度或色彩进行比较(也可以根据检测目的采用像素的纹理特征、植被指数特征进行比较),判断每个像素是否发生变化,进而检测出变化区域。基于像素的影像变化检测易受影像配准、辐射校正等因素的影响。基于特征的影像变化检测需首先确定感兴趣的对象并提取其特征,然后通过特征的比较,获取该对象的变化信息。

The key of remote sensing change detection is to find and extract the change information between remote sensing images through the analysis of different time phases. The methods of image change detection (also referred to the change detection between remote sensing images in different phases) are mainly divided into two categories: one is pixel based image change detection, the other is feature-based image change detection. Pixel based image change detection is to compare the gray or color of the two phases before and after each pixel on the basis of accurate geometric registration (or compare the texture features and vegetation index features of the pixel according to the detection purpose), judge whether each pixel has changed, and then detect the changed area. Pixel based image change detection is easily affected by image registration, radiometric correction and other factors. In feature-based image change detection, firstly, the object of interest is determined and its features are extracted, and then the change information of the object is obtained by comparing the features.

检测内容

Test content

变化检测通常包括以下 4 个方面的内容:

Change detection usually includes the following four aspects:

(1)判断是否发生了变化,即确定研究区域内地物是否发生了变化;

(1) To judge whether there has been a change, that is, to determine whether there has been a change in the study area;

(2)标定变化发生的区域,即确定在何处发生了变化,将变化像元与未变化像元区分开来;

(2) To demarcate the change area, that is, to determine where the change occurs, and to distinguish the changed pixel from the unchanged pixel;

(3)鉴别变化的性质,给出在每个变化像元上所发生变化的类型,即确定变化前后该像元处的地物类型;

(3) The nature of change is identified, and the type of change on each changed pixel is given, that is, the type of surface features at the pixel before and after the change is determined;

(4)评估变化的时间和空间分布模式。

(4) Assess the temporal and spatial patterns of change.

其中,前两个方面是变化检测所要解决的基本问题,而后两个方面则根据应用要求决定是否需要做。

Among them, the first two aspects are the basic problems to be solved by change detection, while the latter two aspects decide whether to do it according to the application requirements.

三个层次

Three levels

遥感图像分析过程中通常包括数据层处理、特征层处理和目标层处理三个过程。依据这三个层次划分,可将变化检测分为:像元级变化检测、特征级变化检测和目标级变化检测。

The process of remote sensing image analysis usually includes data layer processing, feature layer processing and target layer processing. According to these three levels, change detection can be divided into pixel level change detection, feature level change detection and target level change detection.

像元级

Pixel level

像元级变化检测是指直接在采集的原始图像上进行变化检测。尽管基于像元的变化检测有它一定的局限性,但由于它是基于最原始的图像数据,能更多地保留图像原有的真实感,提供其它变化检测层次所不能提供的细微信息,因而目前绝大多数的变化检测方法都是像元级变化检测。

Pixel level change detection refers to change detection directly on the original image. Although pixel based change detection has its limitations, most of the current change detection methods are pixel level change detection because it is based on the most original image data and can retain the original reality of the image and provide more subtle information than other change detection levels can provide.

特征级

Characteristic level

特征级变化检测是采用一定的算法先从原始图像中提取特征信息,如边缘、形状、轮廓、纹理等,然后对这些特征信息进行综合分析与变化检测。由于特征级的变化检测对特征进行关联处理,把特征分类成有意义的组合,因而它对特征属性的判断具有更高的可信度和准确性。但它不是基于原始数据而是特征,所以在特征提取过程中不可避免地会出现信息的部分丢失,难以提供细微信息。

Feature level change detection uses a certain algorithm to extract feature information from the original image, such as edge, shape, contour, texture, etc., and then comprehensively analyze and detect these feature information. Because the feature level change detection processes the features and classifies them into meaningful combinations, it has higher reliability and accuracy in the judgment of feature attributes. But it is not based on the original data, but features, so in the process of feature extraction, part of the information will inevitably be lost, and it is difficult to provide subtle information.

目标级

Target level

目标级变化检测主要检测某些特定对象(比如道路、房屋等具有明确含义的目标)是在图像理解和图像识别的基础上进行的变化检测,它是一种基于目标模型的高层分析方法。

Target level change detection mainly detects some specific objects (such as roads, houses and other objects with clear meaning). It is a change detection based on image understanding and image recognition. It is a high-level analysis method based on target model.

变化检测的三个层次在实现上各有优缺点,在具体的变化检测中究竟检测到哪个层次是根据任务的需要确定的。像元级的变化检测保持了尽可能多的原始信息,具有特征级和目标级层次上所不具备的细节信息,但像元级变化检测仅考虑像素属性的变化,而未考虑其空间等特征属性的变化;特征级变化检测不仅考虑到空间形状的变化,而且还要考虑特征属性的变化,但特征级的变化检测依赖于特征提取的结果,但特征提取本身比较困难;目标级的变化检测最大的优点是它接近用户的需求,检测的结果可直接应用,但它的不足之处在于目标提取的困难性。

The three levels of change detection have their own advantages and disadvantages in the implementation. Which level is detected in the specific change detection is determined according to the needs of the task. The pixel level change detection keeps as much original information as possible, and has the detail information that is not available at the feature level and target level. However, the pixel level change detection only considers the change of the pixel attributes, but not the change of the spatial and other feature attributes. The feature level change detection not only considers the change of the spatial shape, but also considers the change of the feature attributes, but also considers the change of the feature level However, feature extraction itself is difficult; the biggest advantage of target level change detection is that it is close to the needs of users, and the detection results can be directly applied, but its disadvantage lies in the difficulty of target extraction.

检测方法

test method

影像变化检测的方法主要分为两大类:一类是基于像素的影像变化检测,一类是基于特征的影像变化检测。

Image change detection methods are mainly divided into two categories: one is pixel based image change detection, the other is feature-based image change detection.

基于像素

Pixel based

基于像素的影像变化检测方法的优点为:方法简单、速度快,容易获得变化区域,但不能确定影像变化类型和性质。它的具体算法有:差值法、比值法、相关系数法、回归分析法等。

The advantages of pixel based image change detection method are: the method is simple, fast, easy to obtain the change area, but can not determine the type and nature of image change. Its specific algorithms are: difference method, ratio method, correlation coefficient method, regression analysis method, etc.

(1)差值法。先计算前后两个时相遥感影像对应像素灰度值(或色彩值)的差值,生成差值影像,然后对差值影像进行阈值化,就可以检测出变化区域。

(1) Difference method. Firstly, the difference between the gray value (or color value) of the corresponding pixels of the two temporal remote sensing images is calculated to generate the difference image, and then the difference image is thresholded to detect the change area.

(2)比值法。先计算前后两个时相遥感影像对应像素灰度值(或色彩值)的比值,生成比值影像,如果某个像素上没有发生变化,则其比值接近 1,反之比值将明显高于或低于 1.因此只要预先设定合理的低阈值和高阈值,就可以检测出变化区域。

(2) Ratio method. Firstly, the ratio of the gray value (or color value) of the corresponding pixels of the two temporal remote sensing images is calculated to generate a ratio image. If there is no change in a pixel, the ratio is close to 1, otherwise, the ratio will be significantly higher or lower than 1. Therefore, as long as a reasonable low threshold and high threshold are set in advance, the change area can be detected.

(3)相关系数法。先计算前后两个时相遥感影像对应像素灰度的相关系数,如果相关系数值接近 1,说明该像素没有明显变化,反之则说明该像素发生了变化。

(3) Correlation coefficient method. First, calculate the gray correlation coefficient of the corresponding pixel of the two temporal remote sensing images. If the correlation coefficient value is close to 1, it means that the pixel has no obvious change, otherwise, it means that the pixel has changed.

(4)回归分析法。该方法将某一时刻影像的像素灰度值看做另一时刻影像对应像素灰度值的一个线性函数,并用最小二乘法估计此线性函数。由于发生变化的像素将有一个不同于由回归函数预测的灰度值,因此当回归函数预测的灰度值和实际灰度值的差值大于给定的阈值时,就认为该像素发生了变化。

(4) Regression analysis. In this method, the pixel gray value of an image at one time is regarded as a linear function of the corresponding pixel gray value of an image at another time, and the linear function is estimated by the least square method. Because the changed pixel will have a gray value different from that predicted by the regression function, when the difference between the gray value predicted by the regression function and the actual gray value is greater than the given threshold, it is considered that the pixel has changed.

基于特征

Feature based

基于特征的影像变化检测方法的特点为:首先要先确定检测对象,并提取该对象的特征数据,然后通过特征的比较,获取检测对象的变化信息。

The characteristics of feature-based image change detection method are as follows: firstly, the detection object is determined, and the feature data of the object is extracted, and then the change information of the detection object is obtained by comparing the features.

确定检测对象就是要从遥感影像中分离出检测对象,检测对象可能是线状地物(如道路、沟渠等),也可能是面状地物(如水库、湖泊等),还可能是三维空间中的复杂地物(如建筑物)。根据不同类型的检测对象,可采用不同的方法来提取其特征数据。例如在几何特征提取方面,线状地物可采用线检测、细化、跟踪等算法获得其中心线的位置;面状地物可采用区域分割、边界探测等算法获取其边界线数据;三维空间中的复杂地物可采用摄影测量技术获得其轮廓线数据。

To determine the detection object is to separate the detection object from the remote sensing image. The detection object may be linear objects (such as roads, ditches, etc.), surface objects (such as reservoirs, lakes, etc.), or complex objects (such as buildings) in three-dimensional space. According to different types of detection objects, different methods can be used to extract their feature data. For example, in the aspect of geometric feature extraction, line detection, thinning, tracking and other algorithms can be used to obtain the position of the center line of linear objects; area segmentation, boundary detection and other algorithms can be used to obtain the boundary line data of surface objects; photogrammetry technology can be used to obtain the contour line data of complex objects in three-dimensional space.

根据特征描述方法的不同,可采用不同的方法来比较两组特征。具体如下:

According to different feature description methods, different methods can be used to compare two groups of features. The details are as follows:

(1)当采用数值特征来描述检测对象时,可采用统计模式识别的方法来判断两组特征的相似程度及确定检测对象的变化信息;

(1) When the numerical features are used to describe the detected object, the statistical pattern recognition method can be used to judge the similarity of the two groups of features and determine the change information of the detected object;

(2)当采用结构特征来描述检测对象时,可采用结构模式识别的方法判断两组特征的相似程度及确定检测对象的变化信息。

(2) When the structural features are used to describe the detected object, the structural pattern recognition method can be used to judge the similarity of the two groups of features and determine the change information of the detected object.

基于特征的影像变化检测是非常复杂的影像处理、识别和理解的过程,目前还没有实用的成果出现,仍处于探索之中。

Feature based image change detection is a very complex process of image processing, recognition and understanding. At present, there are no practical results and it is still in exploration.

一般流程

General process

遥感影像变化检测的一般处理流程包括图像预处理、变化检测及检测结果输出三个部分。

The general processing flow of remote sensing image change detection includes image preprocessing, change detection and detection result output.

预处理

Pretreatment

由于获取条件的差异,多时相遥感图像中存在非地物变化而造成的图像变化。因此,消除非地物变化是变化检测中不可缺少的步骤,在图像预处理的过程中,最重要的两个过程是几何校正和辐射校正。

Due to the difference of acquisition conditions, there are image changes caused by non object changes in multi temporal remote sensing images. Therefore, the elimination of non surface changes is an indispensable step in change detection. In the process of image preprocessing, the two most important processes are geometric correction and radiometric correction.

变化信息获取

Change information acquisition

变化信息获取是变化检测处理中的关键步骤。在这个过程中,先根据变化检测对象,进行变化特征提取(包括光谱特征及结构特征)。提取的变化特征,有些可以直接用于变化分析称为确定变化特征;有些不能直接用于变化分析,需要结合辅助信息(如在目标级的变化检测中,需要结合目标的模型来描述变化)称为非确定变化特征。最后对变化信息进行分析与描述得到检测结果。

Change information acquisition is a key step in change detection and processing. In this process, the change features (including spectral features and structural features) are extracted according to the change detection object. Some of the extracted change features can be directly used for change analysis, which is called deterministic change features; some can not be directly used for change analysis, which need to be combined with auxiliary information (for example, in the target level change detection, it needs to be combined with the target model to describe the change) is called uncertain change features. Finally, the change information is analyzed and described to get the detection results.

检测结果后处理及输出

Post processing and output of test results

变化检测的后处理是指对得到的检测结果进行再次处理以满足实际需求。主要方法包括滤波处理、数学形态学处理等。变化检测的结果根据用户的需要可以以报表的形式输出,也可以以变化图的形式输出,还可以存储在数据库中。通常情况下,像元级的变化检测是将变化和未变化的区域以二值图的方式进行表示。在特征级和目标级的变化检测中,不仅需要标注出变化的特征或目标,而且需要输出描述特征或目标变化的各种参数。

The post-processing of change detection is to re process the detection results to meet the actual needs. The main methods include filtering, mathematical morphology and so on. The results of change detection can be output in the form of report or change graph according to the needs of users, and can also be stored in the database. In general, pixel level change detection is to represent the changed and unchanged regions in the form of binary image. In feature level and target level change detection, not only the changing features or targets need to be marked, but also various parameters describing the changes of features or targets need to be output.


速搜资源网 , 版权所有丨如未注明 , 均为原创丨转载请注明原文链接:【速搜问答】影像变化检测是什么
喜欢 (0)
[361009623@qq.com]
分享 (0)
发表我的评论
取消评论
表情 贴图 加粗 删除线 居中 斜体 签到

Hi,您需要填写昵称和邮箱!

  • 昵称 (必填)
  • 邮箱 (必填)
  • 网址