技术领域technical field
本发明属于水体遥感影像技术领域，主要涉及一种卫星遥感影像的城市水体提取方法。The invention belongs to the technical field of water body remote sensing images, and mainly relates to a method for extracting urban water bodies from satellite remote sensing images.
背景技术Background technique
城市是人类社会高度发展的体现，城市水体作为城市生态系统中重要的因素，在维持城市生态系统稳定性上具有十分重要的作用。城市水体的改变将会对生活产生巨大的变化，可能会引发一些灾害，如城市水源滞留、水资源短缺，甚至引发一些和人类健康生活相关的疾病。因此，了解和掌握城市水体分布以及水域面积的变化已经成为了人们日益关注的焦点。Cities are the embodiment of the high development of human society. As an important factor in urban ecosystems, urban water bodies play a very important role in maintaining the stability of urban ecosystems. Changes in urban water bodies will have huge changes in life, and may cause some disasters, such as urban water retention, water shortages, and even some diseases related to human health and life. Therefore, understanding and mastering the distribution of urban water bodies and the change of water area has become the focus of people's increasing attention.
近年来，随着遥感技术的应用与发展，遥感影像在自然资源调查、动态监测、自然地表水源规划等方面发挥着着越来越重要的角色，利用遥感技术进行地表的监测也得到了越来越多的科研工作者的关注；遥感影像可以以一个不同的视角去观察地球表面的地物，实时的监测地表的变化。在水体提取技术中，通过遥感数据及时准确地获取城市水体信息成为目前主流的水体提取方式。到目前为止，诸多学者提出了大量遥感影像的水体提取方法，但大部分都是基于中低分辨率遥感影像，由于影像分辨率较低，面积较小的水体未能有效的提取；尤其是对于城市地区水体的提取，城市地区水体面积大小参差不齐，存在诸多小面积人工湖泊和细小的河流。因此，在城市水体提取中应更多的采用高分辨率遥感影像，以资源3号卫星为例，资源3号卫星遥感影像有着5.8m分辨率和5.2km的宽幅，它为城市水体提取提供了理想的多光谱影像数据，资源3号具体参数如下表1：In recent years, with the application and development of remote sensing technology, remote sensing images have played an increasingly important role in natural resource investigation, dynamic monitoring, and natural surface water source planning. More and more scientific researchers pay attention; remote sensing images can observe the features on the earth's surface from a different perspective, and monitor the changes of the earth's surface in real time. In water body extraction technology, timely and accurate acquisition of urban water body information through remote sensing data has become the current mainstream water body extraction method. So far, many scholars have proposed a large number of water body extraction methods for remote sensing images, but most of them are based on low-resolution remote sensing images. Due to the low resolution of images, small water bodies cannot be effectively extracted; especially for Extraction of water bodies in urban areas, the size of water bodies in urban areas is uneven, and there are many small artificial lakes and small rivers. Therefore, more high-resolution remote sensing images should be used in the extraction of urban water bodies. Taking the Ziyuan 3 satellite as an example, the remote sensing images of the Ziyuan 3 satellite have a resolution of 5.8m and a width of 5.2km. To obtain the ideal multispectral image data, the specific parameters of Resource No. 3 are shown in Table 1:
表1资源三号卫星影像参数Table 1 Image parameters of ZY-3 satellite
水体是遥感影像中很常见的地物类别，也是重要的基础地理信息之一，其动态信息的快速获取，对水资源调查、水利规划、环境监测与保护等事业都有着十分明显的实用价值和科学意义。对此，诸多学者很早就对此展开了研究，提出了许多有效的水体信息自动提取模型。大致可以分为4类：(a)单波段和多波段阈值分割法(single-band or multiple-band threshold method)，(b)水体指数法(Water indices)，(c)线性分解模型(linearun-mixing model)，(d)监督与非监督分类方法(supervised or unsupervisedclassification method)。除此之外，还有一些其他的方法，如：基于数字高程模型的水体提取技术，基于微波遥感(Microwave Remote Sensing)影像的水体提取技术，面向对象(Object Oriented)技术的水体提取技术等。但这些方法并不常用，总得来说，水体指数法由于其模型简单、方便，且精度较高，在实际中最为常用。Water body is a very common feature category in remote sensing images, and it is also one of the important basic geographic information. The rapid acquisition of its dynamic information has very obvious practical value and scientific significance. In this regard, many scholars have carried out research on this for a long time, and proposed many effective models for automatic extraction of water body information. It can be roughly divided into four categories: (a) single-band or multiple-band threshold method (single-band or multiple-band threshold method), (b) water index method (Water indices), (c) linear decomposition model (linearun- mixing model), (d) supervised and unsupervised classification method (supervised or unsupervised classification method). In addition, there are some other methods, such as: water body extraction technology based on digital elevation model, water body extraction technology based on microwave remote sensing (Microwave Remote Sensing) image, water body extraction technology based on Object Oriented technology, etc. However, these methods are not commonly used. Generally speaking, the water body index method is most commonly used in practice because of its simple model, convenience, and high accuracy.
然而随着遥感影像分辨率的提高，大部分高分辨率遥感影像(如；WorldView-2,IKONOS,RapidEye,and资源3号)没有像Landsat TM/ETM+/OLI这么多可以利用的波段用于水体的提取，因此，MNDWI(改进的归一化差异水体指数法)和AWEI(自动水体指数)将无法使用，因为大部分的高分辨率遥感影像只有4个波段(蓝色、绿色、红色、近红外波段)，缺少MNDWI/AWEI计算所需的短红外波段(SWIR)。因此，在使用NDWI(归一化差异水体指数)对高分辨率影像进行水体提取的时候就将会产生一些问题，如阴影无法去除的问题，尤其是城市地区高层建筑物的阴影，在高分辨率影像上表现尤为突出。在高分辨率遥感影像上，城市区域的高层建筑物阴影与水体很难区分，尽管目前有相关学者对这一方面进行了研究，如基于面向对象的技术，通过计算高层建筑物阴影区域的纹理特性来对高层建筑物阴影进行检测；虽然可以达到预期效果，但由于纹理的描述与计算相对复杂且耗时较长，所以从计算时间上考虑该方法并不是一个理想的阴影检测方法。也有基于SVM特征训练进行的阴影检测，以期达到去除水体检测中高层建筑物阴影的影响。SVM是一种分类精度较高的方法，但SVM训练需要花去较多的时间，尤其是当训练样本数目较多且样本特征向量维度较高时。若通过采用高分辨率遥感影像阴影检测方法(morphological shadow index,MSI)与NDWI相结合的方式对WorldView-2高分辨率影像进行水体提取，以期提高水体检测精度；尽管该方法原理简单，但由于该方法以NDWI方法作为水体提取的基础，其检测精度并不会很高，尤其是周围植被茂密的细小面积水域，水体的光谱特性将受到严重的污染，水体光谱特征表现极其不稳定，同时城区水体具有悬浮泥沙含量高、水体富营养化严重、受各种污染物污染较大等特点，使得城区水体与自然界中未受污染的水体表现不同的光学特性。However, with the improvement of remote sensing image resolution, most high-resolution remote sensing images (such as; WorldView-2, IKONOS, RapidEye, and Resource No. 3) do not have as many available bands as Landsat TM/ETM+/OLI for water bodies Therefore, MNDWI (Modified Normalized Difference Water Index) and AWEI (Automatic Water Index) will not be available, because most high-resolution remote sensing images only have 4 bands (blue, green, red, near infrared band), lacking the short infrared band (SWIR) required for MNDWI/AWEI calculations. Therefore, when using NDWI (Normalized Difference Water Index) to extract water from high-resolution images, there will be some problems, such as the problem that shadows cannot be removed, especially the shadows of high-rise buildings in urban areas. The performance is particularly prominent on the rate image. On high-resolution remote sensing images, the shadows of high-rise buildings in urban areas are difficult to distinguish from water bodies, although some scholars have conducted research on this aspect, such as based on object-oriented technology, by calculating the texture of shadow areas of high-rise buildings characteristics to detect the shadows of high-rise buildings; although the expected effect can be achieved, due to the relatively complex and time-consuming texture description and calculation, this method is not an ideal shadow detection method in terms of calculation time. There is also shadow detection based on SVM feature training in order to remove the influence of high-rise building shadows in water body detection. SVM is a method with high classification accuracy, but SVM training takes more time, especially when the number of training samples is large and the dimension of the sample feature vector is high. If the combination of high-resolution remote sensing image shadow detection method (morphological shadow index, MSI) and NDWI is used to extract water bodies from WorldView-2 high-resolution images, in order to improve the accuracy of water body detection; although the principle of this method is simple, due to This method uses the NDWI method as the basis for water body extraction, and its detection accuracy is not very high, especially in the small area of water with dense vegetation around, the spectral characteristics of the water body will be seriously polluted, and the spectral characteristics of the water body are extremely unstable. The water body has the characteristics of high content of suspended sediment, serious eutrophication of the water body, and high pollution by various pollutants, which makes the urban water body and the unpolluted water body in nature show different optical characteristics.
发明内容Contents of the invention
针对如何从高分辨率卫星遥感影像上，进行城市区域的水体提取问题，特别是有效的区分高层建筑物阴影与水体和提高水体提取的精度问题，本发明提出了一种卫星遥感影像的城市水体提取方法(Automatic urban water extraction method,AUWEM)。Aiming at the problem of how to extract water bodies in urban areas from high-resolution satellite remote sensing images, especially the problem of effectively distinguishing the shadows of high-rise buildings from water bodies and improving the accuracy of water body extraction, the present invention proposes a satellite remote sensing image of urban water bodies Extraction method (Automatic urban water extraction method, AUWEM).
本发明的具体技术方案如下：Concrete technical scheme of the present invention is as follows:
一种卫星遥感影像的城市水体提取方法，包括以下步骤：A method for extracting urban water bodies from satellite remote sensing images, comprising the following steps:
步骤1：遥感影像数据的预处理，即对遥感影像数据进行正射校正和大气校正；Step 1: Preprocessing of remote sensing image data, that is, performing orthorectification and atmospheric correction on remote sensing image data;
步骤2：预处理后的遥感影像数据，包括蓝色band1数据、绿色band2数据、红色band3数据和近红外band4数据，选取预处理后的遥感影像数据中的蓝色band1数据代替归一化差异水体指数NDWI的计算公式中的绿色band2数据，获得新的归一化差异水体指数NNDWI1，新的归一化差异水体指数NNDWI1的计算公式为：Step 2: Preprocessed remote sensing image data, including blue band1 data, green band2 data, red band3 data and near-infrared band4 data, select the blue band1 data in the preprocessed remote sensing image data to replace the normalized difference water body The green band2 data in the calculation formula of index NDWI can obtain the new normalized difference water index NNDWI1, and the calculation formula of the new normalized difference water index NNDWI1 is:
此计算公式即为NNDWI1指数模型，利用此模型通过阈值分割即得到NNDWI1的阈值分割结果，也即为NNDWI1水体提取结果；This calculation formula is the NNDWI1 index model. Using this model, the threshold segmentation result of NNDWI1 can be obtained through threshold segmentation, which is the result of NNDWI1 water body extraction;
步骤3：对预处理后的遥感影像数据中包括的四个波段数据，即蓝色band1数据、绿色band2数据、红色band3数据和近红外band4数据进行PCA变换，并将PCA变换后的第一主成分分量Component1替代归一化差异水体指数NDWI的计算公式中的绿色band2数据，获得另一个新的归一化差异水体指数NNDWI2，即：Step 3: Perform PCA transformation on the four band data included in the preprocessed remote sensing image data, that is, blue band1 data, green band2 data, red band3 data and near-infrared band4 data, and convert the first primary data after PCA transformation The component component Component1 replaces the green band2 data in the calculation formula of the normalized difference water index NDWI to obtain another new normalized difference water index NNDWI2, namely:
其中，Component1表示PCA变换的第一主成分分量，此计算公式即为NNDWI2指数模型，利用此模型通过阈值分割即得到NNDWI2的阈值分割结果，也即为NNDWI2的水体提取结果；Among them, Component1 represents the first principal component component of PCA transformation. This calculation formula is the NNDWI2 index model. Using this model, the threshold segmentation result of NNDWI2 can be obtained through threshold segmentation, which is also the water body extraction result of NNDWI2;
步骤4：将步骤2得到的NNDWI1的阈值分割结果和步骤3中得到的NNDWI2的阈值分割结果进行叠加，将得到的结果定义为新的归一化差异水体指数NNDWI的阈值分割结果，即NNDWI1的阈值分割结果与NNDWI2的阈值分割结果进行叠加，其计算公式为：Step 4: Superimpose the threshold segmentation result of NNDWI1 obtained in step 2 and the threshold segmentation result of NNDWI2 obtained in step 3, and define the obtained result as the threshold segmentation result of the new normalized difference water index NNDWI, that is, the threshold segmentation result of NNDWI1 The threshold segmentation results are superimposed with the threshold segmentation results of NNDWI2, and the calculation formula is:
NNDWI＝(segmentation_NNDWI1)∪(segmentation_NNDWI2)NNDWI＝(segmentation_NNDWI1)∪(segmentation_NNDWI2)
式中segmentation_NNDWI1表示NNDWI1的阈值分割结果，segmentation_NNDWI2表示NNDWI2的阈值分割结果，此计算公式即为NNDWI指数模型，利用此指数模型得到NNDWI的水体提取结果；In the formula, segmentation_NNDWI1 represents the threshold segmentation result of NNDWI1, and segmentation_NNDWI2 represents the threshold segmentation result of NNDWI2. This calculation formula is the NNDWI index model, and the water body extraction result of NNDWI is obtained by using this index model;
步骤5：对预处理后的遥感影像数据中的近红外band4数据进行阈值分割，得到近红外band4数据的阈值分割结果；Step 5: Perform threshold segmentation on the near-infrared band4 data in the preprocessed remote sensing image data, and obtain the threshold segmentation result of the near-infrared band4 data;
步骤6：对NNDWI的水体提取结果中的大面积水体对象和小面积对象进行分割，NNDWI的水体提取结果中，像素个数大于设定阈值的为大面积水体对象，像素个数小于等于设定阈值的为小面积对象；Step 6: Segment large-area water objects and small-area objects in the water body extraction results of NNDWI. In the water body extraction results of NNDWI, those with the number of pixels greater than the set threshold are large-area water objects, and the number of pixels is less than or equal to the set value. The threshold is for small area objects;
步骤7：对步骤6中得到的小面积对象进行数学形态学膨胀处理，得到膨胀后的小面积对象，将步骤5得到的近红外band4数据的阈值分割结果作为约束条件，即采用膨胀后的小面积对象和近红外band4数据的阈值分割结果求交集的方式对膨胀后的小面积对象进行约束，约束的数学表达式为：Step 7: Perform mathematical morphology expansion on the small-area object obtained in step 6 to obtain the expanded small-area object, and use the threshold segmentation result of the near-infrared band4 data obtained in step 5 as a constraint condition, that is, use the expanded small-area object The intersection of the threshold segmentation results of the area object and the near-infrared band4 data is used to constrain the expanded small-area object. The mathematical expression of the constraint is:
component2＝(dilate_component)∩(segmentation_band4)component2=(dilate_component)∩(segmentation_band4)
式中，dilate_component表示膨胀后的小面积对象，segmentation_band4表示近红外band4数据的阈值分割结果，component2表示约束后的小面积对象；In the formula, dilate_component represents the small-area object after dilation, segmentation_band4 represents the threshold segmentation result of near-infrared band4 data, and component2 represents the constrained small-area object;
步骤8：对步骤7得到的约束后的小面积对象进行阴影检测与去除，得到小面积水体对象；Step 8: Perform shadow detection and removal on the constrained small-area object obtained in step 7 to obtain a small-area water body object;
阴影检测与去除，是指对每个小面积对象中的每个像元进行波谱关系的描述，并判断该像元是否满足阴影像元的条件，记录并统计每个小面积对象中阴影像元的个数，当一个小面积对象中阴影像元所占比例大于阈值T时，把该小面积对象判定为建筑物阴影对象，小于等于阈值T时小面积对象则判定为小面积水体对象，阴影像元所占比例即为小面积对象中阴影像元的个数与该小面积对象中总像元个数的比值，区分小面积对象中小面积水体对象和阴影对象的函数表达式为：Shadow detection and removal refers to describing the spectral relationship of each pixel in each small-area object, and judging whether the pixel meets the conditions of shadow pixels, recording and counting the shadow pixels in each small-area object When the proportion of shadow pixels in a small-area object is greater than the threshold T, the small-area object is judged as a building shadow object; when it is less than or equal to the threshold T, the small-area object is judged as a small-area water body object, and the shadow The proportion of the pixel is the ratio of the number of shadow pixels in the small-area object to the total number of pixels in the small-area object. The function expression for distinguishing the small-area water object and the shadow object in the small-area object is:
式中，n表示某一小面积对象中总像元个数，m为该小面积对象中阴影像元的个数；In the formula, n represents the total number of pixels in a small-area object, and m is the number of shadow pixels in the small-area object;
阴影像元的条件，是指满足阴影像元的波谱大小关系，即满足以下三个不等式之一：The condition of the shaded pixel refers to satisfying the spectral size relationship of the shaded pixel, that is, satisfying one of the following three inequalities:
步骤9：将步骤6中得到的大面积水体对象与步骤8中得到的小面积水体对象进行叠加，即将步骤6中得到的大面积水体对象和步骤8中得到的小面积水体对象求并集，得到卫星遥感影像的城市水体提取结果。Step 9: Superimpose the large-area water object obtained in step 6 with the small-area water object obtained in step 8, that is, the union of the large-area water object obtained in step 6 and the small-area water object obtained in step 8, The extraction results of urban water bodies from satellite remote sensing images are obtained.
本发明的有益效果如下：The beneficial effects of the present invention are as follows:
(1)新的归一化差异水体指数NNDWI1和NNDWI2可以有效的对水体进行初始提取，提高后续水体的提取精度。(1) The new normalized difference water index NNDWI1 and NNDWI2 can effectively extract the initial water body and improve the accuracy of subsequent water body extraction.
(2)在阴影对象获取方面，为了更加准确的获取阴影对象，对小面积对象进行膨胀处理；同时为了限制在真实地表阴影区域，对膨胀的对象采用近红外band4数据的阈值分割结果进行约束。(2) In terms of shadow object acquisition, in order to obtain shadow objects more accurately, small-area objects are expanded; at the same time, in order to limit the real surface shadow area, the threshold segmentation results of near-infrared band4 data are used to constrain the expanded objects.
(3)为避免阴影对象检测过程中时间消耗，对小面积对象采用基于波谱特性的描述，减少传统中采用纹理特征进行描述的时间消耗，提高方法的计算效率。(3) In order to avoid time consumption in the shadow object detection process, the description based on spectral characteristics is used for small-area objects, which reduces the time consumption of the traditional texture feature description and improves the computational efficiency of the method.
(4)AUWEM方法的分类精度要高于NDWI的分类精度和最大似然法的分类精度。在5个实验区的实验结果中，AUWEM平均Kappa系数约为93.0％，NDWI的平均Kappa系数约为86.2％，最大似然法分类精度介于两者之间，平均Kappa系数约为88.6％；同时AUWEM漏检与虚警总错误率也要少于NDWI的分类结果和最大似然法的分类结果，AUWEM平均漏检与虚警总错误率约为11.9％，最大似然法平均漏检与虚警总错误率约为18.2％，NDWI平均漏检与虚警总错误率约为22.1％％。除此之外，AUWEM具有更高的水体边缘检测精度。(4) The classification accuracy of AUWEM method is higher than that of NDWI and the classification accuracy of maximum likelihood method. In the experimental results of the five experimental areas, the average Kappa coefficient of AUWEM is about 93.0%, the average Kappa coefficient of NDWI is about 86.2%, and the classification accuracy of the maximum likelihood method is between the two, and the average Kappa coefficient is about 88.6%; At the same time, the total error rate of missed detection and false alarm of AUWEM is also less than the classification results of NDWI and the classification result of maximum likelihood method. The average total error rate of missed detection and false alarm of AUWEM is about 11.9%. The total error rate of false alarm is about 18.2%, and the average error rate of missed detection and false alarm of NDWI is about 22.1%. Besides, AUWEM has higher accuracy of water body edge detection.
附图说明Description of drawings
图1是本发明具体实施方式中的NNDWI的水体提取结果，其中图(a)是从资源3号卫星获得的遥感影像，图(b)是NNDWI1水体提取结果，图(c)是NNDWI2水体提取结果，图(d)NNDWI水体提取结果；Fig. 1 is the water body extraction result of NNDWI in the specific embodiment of the present invention, wherein figure (a) is the remote sensing image that obtains from No. 3 resource satellite, figure (b) is NNDWI1 water body extraction result, figure (c) is NNDWI2 water body extraction Results, Figure (d) NNDWI water body extraction results;
图2是本发明具体实施方式中的膨胀约束过程示意图；Fig. 2 is a schematic diagram of the expansion constraint process in a specific embodiment of the present invention;
图3(a)～(e)是本发明具体实施方式中的阴影区域不同波谱特性曲线；Fig. 3 (a)～(e) are different spectrum characteristic curves of the shaded area in the specific embodiment of the present invention;
图4是本发明具体实施方式中的卫星遥感影像的城市水体提取方法流程图；Fig. 4 is the flow chart of the urban water body extraction method of the satellite remote sensing image in the specific embodiment of the present invention;
图5是本发明具体实施方式中的不同方法在5个实验区的实验结果；Fig. 5 is the experimental result of different methods in 5 experimental areas in the specific embodiment of the present invention;
图6(a)～(f)是本发明具体实施方式中的不同方法在5个实验区的6个指标直方图；Fig. 6 (a)～(f) are 6 indicator histograms in 5 experimental areas of different methods in the specific embodiment of the present invention;
图7是本发明具体实施方式中的水体边缘精度评估中待评估区域获取示意图；Fig. 7 is a schematic diagram of acquisition of the area to be evaluated in the evaluation of the edge accuracy of the water body in the specific embodiment of the present invention;
图8(a)～(c)是本发明具体实施方式中的不同方法在5个实验区水体边缘检测精度比较。Fig. 8(a)-(c) are comparisons of detection accuracy of water body edge in 5 experimental areas by different methods in specific embodiments of the present invention.
具体实施方式Detailed ways
下面结合附图和具体实施方式对本发明进行详细说明。The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments.
一种卫星遥感影像的城市水体提取方法，以资源3号卫星遥感影像数据为例，包括以下步骤：A method for extracting urban water bodies from satellite remote sensing images, taking Ziyuan No. 3 satellite remote sensing image data as an example, comprising the following steps:
步骤1：遥感影像数据的预处理，即对遥感影像数据进行正射校正和大气校正；对实验区域的影像利用RPC+30DEM进行无控制点正射校正，采用Feyisa G L et al.的FLAASH(Fast Line-of-SightAtmosphericAnalysis ofSpectral Hypercubes)大气校正模型进行影像的大气校正，均在ENVI5.2软件中完成。其中，资源3号FLAASH大气校正需要的定标系数可通过网站http://www.cresda.com/CN/Downloads/dbcs/index.shtml进行下载，光谱响应函数可通过网站http://www.cresda.com/CN/Downloads/gpxyhs/index.shtml进行下载。Step 1: Preprocessing of remote sensing image data, that is, performing orthorectification and atmospheric correction on the remote sensing image data; using RPC+30DEM to perform orthorectification without control points on the image of the experimental area, using Feyisa G L et al.’s FLAASH (Fast Line-of-Sight Atmospheric Analysis of Spectral Hypercubes) atmospheric correction model for image atmospheric correction, all completed in ENVI5.2 software. Among them, the calibration coefficients required for the FLAASH atmospheric correction of Resource No. 3 can be downloaded through the website http://www.cresda.com/CN/Downloads/dbcs/index.shtml, and the spectral response function can be downloaded through the website http://www. cresda.com/CN/Downloads/gpxyhs/index.shtml to download.
步骤2：预处理后的遥感影像数据，包括蓝色band1数据、绿色band2数据、红色band3数据和近红外band4数据，选取预处理后的遥感影像数据中的蓝色band1数据代替归一化差异水体指数NDWI(NormalizedDifferenceWater Index)的计算公式中的绿色band2数据，获得新的归一化差异水体指数NNDWI1(New NormalizedDifferenceWaterIndex 1)，新的归一化差异水体指数NNDWI1的计算公式为：Step 2: Preprocessed remote sensing image data, including blue band1 data, green band2 data, red band3 data and near-infrared band4 data, select the blue band1 data in the preprocessed remote sensing image data to replace the normalized difference water body The green band2 data in the calculation formula of the index NDWI (Normalized Difference Water Index) can obtain the new normalized difference water index NNDWI1 (New Normalized Difference Water Index 1), and the calculation formula of the new normalized difference water index NNDWI1 is:
此表达式即为NNDWI1指数模型。利用此模型在编程软件Matlab 2014a中通过阈值分割得到NNDWI1阈值分割结果，也即为NNDWI1水体提取结果。This expression is the NNDWI1 exponential model. Using this model in the programming software Matlab 2014a, the threshold segmentation result of NNDWI1 is obtained through threshold segmentation, that is, the NNDWI1 water body extraction result.
步骤3：对预处理后的遥感影像数据中包括的四个波段数据，即蓝色band1数据、绿色band2数据、红色band3数据和近红外band4数据进行PCA变换，并将PCA变换后的第一主成分分量Component1替代归一化差异水体指数NDWI的计算公式中的绿色band2数据，获得另一个新的归一化差异水体指数NNDWI2(New NormalizedDifferenceWaterIndex 2)，即：Step 3: Perform PCA transformation on the four band data included in the preprocessed remote sensing image data, that is, blue band1 data, green band2 data, red band3 data and near-infrared band4 data, and convert the first primary data after PCA transformation The component component Component1 replaces the green band2 data in the calculation formula of the normalized difference water index NDWI to obtain another new normalized difference water index NNDWI2 (New NormalizedDifferenceWaterIndex 2), namely:
其中，Component1表示PCA变换的第一主成分分量。Among them, Component1 represents the first principal component component of PCA transformation.
此表达式即为NNDWI2指数模型。利用此模型在编程软件Matlab 2014a中通过阈值分割得到NNDWI2阈值分割结果，也即为NNDWI2水体提取结果。This expression is the NNDWI2 exponential model. Using this model in the programming software Matlab 2014a, the NNDWI2 threshold segmentation result is obtained through threshold segmentation, which is also the NNDWI2 water body extraction result.
步骤4：将步骤2得到的NNDWI1的阈值分割结果和步骤3中得到的NNDWI2的阈值分割结果进行叠加，将得到的结果定义为新的归一化差异水体指数NNDWI(New NormalizedDifferenceWaterIndex)的阈值分割结果，即NNDWI1的阈值分割结果与NNDWI2的阈值分割结果进行叠加，其计算公式为：Step 4: Superimpose the threshold segmentation result of NNDWI1 obtained in step 2 and the threshold segmentation result of NNDWI2 obtained in step 3, and define the obtained result as the threshold segmentation result of the new normalized difference water index NNDWI (New NormalizedDifferenceWaterIndex) , that is, the threshold segmentation results of NNDWI1 and the threshold segmentation results of NNDWI2 are superimposed, and the calculation formula is:
NNDWI＝(segmentation_NNDWI1)∪(segmentation_NNDWI2)NNDWI＝(segmentation_NNDWI1)∪(segmentation_NNDWI2)
式中segmentation_NNDWI1表示NNDWI1指数影像的阈值分割结果，segmentation_NNDWI2表示NNDWI2指数影像的阈值分割结果。此计算公式即为NNDWI指数模型，利用此模型在编程软件Matlab 2014a中得到NNDWI水体提取结果。In the formula, segmentation_NNDWI1 represents the threshold segmentation result of the NNDWI1 index image, and segmentation_NNDWI2 represents the threshold segmentation result of the NNDWI2 index image. This calculation formula is the NNDWI index model, which is used to obtain the NNDWI water body extraction results in the programming software Matlab 2014a.
通过实验分析，实验中NNDWI1的阈值分割结果所采用的阈值与NNDWI2的阈值分割结果所采用的阈值都设为0，可获得较为理想的实验结果。NNDWI的水体提取结果综合了NNDWI1的水体提取结果和NNDWI2的水体提取结果，避免了单一指数对水体的漏检现象发生。如图1(a)～(d)所示。NNDWI1对含泥沙高的水体检测效果较为明显，而NNDWI2对受植被光谱信息干扰严重的水体较为敏感，因此，实际水体提取中结合NNDWI1的水体提取结果和NNDWI2的水体提取结果，整合生成新的水体提取结果，提高后续水体提取精度。Through experimental analysis, the thresholds used in the threshold segmentation results of NNDWI1 and the threshold segmentation results of NNDWI2 are both set to 0 in the experiment, and ideal experimental results can be obtained. The water body extraction results of NNDWI are integrated with the water body extraction results of NNDWI1 and NNDWI2, which avoids the omission of water bodies by a single index. Figure 1 (a) ~ (d) shown. NNDWI1 is more effective in detecting water bodies with high sediment content, while NNDWI2 is more sensitive to water bodies that are severely interfered by vegetation spectral information. Therefore, in the actual water body extraction, the water body extraction results of NNDWI1 and NNDWI2 are integrated to generate a new The results of water body extraction can improve the accuracy of subsequent water body extraction.
步骤5：对预处理后的遥感影像数据中的近红外band4数据进行阈值分割，得到近红外band4数据的阈值分割结果。Step 5: Perform threshold segmentation on the near-infrared band4 data in the preprocessed remote sensing image data, and obtain the threshold segmentation result of the near-infrared band4 data.
通过分析大量NNDWI水体提取结果的影像资料发现，除部分城区面积较小的人工池塘与湖泊之外，在一般情况下，城市高层建筑物阴影对象的面积一般要小于水体对象的面积。所以在实际的分析中，只对面积相对较小的地物进行检测，因为在这部分面积相对较小的对象中包括几乎所有的阴影对象，同时也包括小面积水体对象。By analyzing a large number of image data of NNDWI water body extraction results, it is found that, except for small artificial ponds and lakes in some urban areas, in general, the area of shadow objects of urban high-rise buildings is generally smaller than that of water objects. Therefore, in the actual analysis, only the objects with a relatively small area are detected, because these objects with a relatively small area include almost all shadow objects, and also include small-area water objects.
步骤6：对NNDWI的水体提取结果中的大面积水体对象和小面积对象进行分割，NNDWI的水体提取结果中，像素个数大于设定阈值的为大面积水体对象，像素个数小于等于设定阈值的为小面积对象；Step 6: Segment large-area water objects and small-area objects in the water body extraction results of NNDWI. In the water body extraction results of NNDWI, those with the number of pixels greater than the set threshold are large-area water objects, and the number of pixels is less than or equal to the set value. The threshold is for small area objects;
小面积对象获取模型，可表示为：The small-area object acquisition model can be expressed as:
其中，t为设定的阈值，其值取最大阴影对象的像素个数，component表示NNDWI水体指数提取结果中的离散对象，area(component)表示NNDWI水体指数提取结果中的离散对象的面积，area(component)>t为大面积水体对象，area(component)≤t为小面积水体或者阴影对象。Among them, t is the set threshold, its value is the number of pixels of the largest shaded object, component represents the discrete object in the NNDWI water body index extraction result, area(component) represents the area of the discrete object in the NNDWI water body index extraction result, area (component)>t is a large-area water body object, and area(component)≤t is a small-area water body or shadow object.
步骤7：对步骤6中得到的小面积对象进行数学形态学膨胀处理，得到膨胀后的小面积对象，将步骤5得到的近红外band4数据的阈值分割结果作为约束条件，即采用膨胀后的小面积对象和近红外band4数据的阈值分割结果求交集的方式对膨胀后的小面积对象进行约束，约束的数学表达式为：Step 7: Perform mathematical morphology expansion on the small-area object obtained in step 6 to obtain the expanded small-area object, and use the threshold segmentation result of the near-infrared band4 data obtained in step 5 as a constraint condition, that is, use the expanded small-area object The intersection of the threshold segmentation results of the area object and the near-infrared band4 data is used to constrain the expanded small-area object. The mathematical expression of the constraint is:
component2＝(dilate_component)∩(segmentation_band4)component2=(dilate_component)∩(segmentation_band4)
式中，dilate_component表示膨胀后的小面积对象，segmentation_band4表示近红外band4数据的阈值分割结果，component2表示约束后的小面积对象；In the formula, dilate_component represents the small-area object after dilation, segmentation_band4 represents the threshold segmentation result of near-infrared band4 data, and component2 represents the constrained small-area object;
该膨胀后的小面积对象更加完整的包括了小面积水体对象与阴影的阴影像素，对膨胀后的小面积对象进行约束也保持了小面积对象的膨胀操作限制在地表真实阴影区域的范围内。The expanded small-area object more completely includes the shadow pixels of the small-area water object and the shadow, and the constraints on the expanded small-area object also keep the expansion operation of the small-area object limited to the range of the real shadow area on the ground surface.
以步骤6得到的小面积对象为建筑物阴影对象为例的膨胀约束过程示意图如图2所示，从资源3号假彩色影像中，分别获取建筑物阴影对象和Band4影像数据阈值分割结果，将得到的建筑物阴影对象进行数学形态学膨胀处理，获得建筑物阴影对象的膨胀结果，将建筑物阴影对象的膨胀结果与Band4影像数据阈值分割结果进行求交集处理，即得到在Band4影像数据阈值分割结果约束下的膨胀后的建筑物阴影对象。Taking the small-area object obtained in step 6 as a building shadow object as an example, the schematic diagram of the expansion constraint process is shown in Figure 2. From the No. The obtained building shadow object is subjected to mathematical morphology expansion processing to obtain the expansion result of the building shadow object, and the expansion result of the building shadow object is intersected with the Band4 image data threshold segmentation result, that is, the Band4 image data threshold segmentation result is obtained. The dilated building shadow object under the resulting constraints.
步骤8：对步骤7得到的约束后的小面积对象进行阴影检测与去除，得到小面积水体对象；Step 8: Perform shadow detection and removal on the constrained small-area object obtained in step 7 to obtain a small-area water body object;
在NNDWI水体提取结果中，基本上只包含小面积水体对象与阴影，所以只需要对小面积水体对象与阴影的特征进行研究与分析，找到适合区分小面积水体对象与阴影的特征。在实验中发现，尽管纹理特征可以很好的描述小面积水体对象与阴影，但由于地物的纹理特征，如：灰度共生矩阵，其计算相对复杂、耗时较长，不太适于用于小面积水体对象与阴影的区分，所以在实验中采用地物的光谱特征描述小面积水体对象和阴影的像素，以此作为依据区分小面积水体对象与阴影。In the NNDWI water body extraction results, basically only small-area water body objects and shadows are included, so it is only necessary to study and analyze the characteristics of small-area water body objects and shadows, and find features suitable for distinguishing small-area water body objects and shadows. In the experiment, it was found that although texture features can well describe small-area water objects and shadows, due to the texture features of ground objects, such as: gray-scale co-occurrence matrix, its calculation is relatively complicated and time-consuming, which is not suitable for Because of the distinction between small-area water objects and shadows, the spectral characteristics of ground objects are used to describe the pixels of small-area water objects and shadows in the experiment, and this is used as a basis to distinguish small-area water objects and shadows.
通过分析大量的水体与阴影的波谱特性曲线，得到水体像元的波谱关系满足不等式：By analyzing the spectral characteristic curves of a large number of water bodies and shadows, it is obtained that the spectral relationship of water body pixels satisfies the inequality:
band2>band4band2>band4
而阴影的像元波谱曲线较为复杂，实验中分析总结出了以下5种波谱特性曲线，如图3(a)～(e)所示。The spectral curve of the shadow pixel is more complicated, and the following five spectral characteristic curves are analyzed and summarized in the experiment, as shown in Figure 3(a)~(e).
根据上述波谱曲线各波段对应的大小关系，实验中总结出阴影像元的波谱关系满足以下三个不等式条件之一：According to the size relationship corresponding to each band of the above spectral curve, the experiment concluded that the spectral relationship of the shadow pixel satisfies one of the following three inequality conditions:
由此可知，阴影像元的条件，是指阴影像元的波谱关系满足以上三个不等式条件。It can be seen that the condition of the shadow pixel means that the spectral relationship of the shadow pixel satisfies the above three inequalities.
阴影检测与去除，是指对每个小面积对象中的每个像元进行波谱关系的描述，并判断该像元是否满足阴影像元的条件，记录并统计每个小面积对象中阴影像元的个数，当一个小面积对象中阴影像元所占比例大于阈值T时，把该小面积对象判定为建筑物阴影对象，小于等于阈值T时小面积对象则判定为小面积水体对象，阴影像元所占比例即为小面积对象中阴影像元的个数与该小面积对象中总像元个数的比值，区分小面积对象中小面积水体对象和阴影对象的函数表达式为：Shadow detection and removal refers to describing the spectral relationship of each pixel in each small-area object, and judging whether the pixel meets the conditions of shadow pixels, recording and counting the shadow pixels in each small-area object When the proportion of shadow pixels in a small-area object is greater than the threshold T, the small-area object is judged as a building shadow object; when it is less than or equal to the threshold T, the small-area object is judged as a small-area water body object, and the shadow The proportion of the pixel is the ratio of the number of shadow pixels in the small-area object to the total number of pixels in the small-area object. The function expression for distinguishing the small-area water object and the shadow object in the small-area object is:
式中，n表示某一小面积对象中总像元个数，m为该小面积对象中阴影像元的个数。阈值T是通过实验统计得来的，对资源3号遥感影像数据的阴影像元进行统计发现，当T取0.5时可以很好的区分水体与阴影对象。In the formula, n represents the total number of pixels in a small-area object, and m is the number of shadow pixels in the small-area object. The threshold T is obtained through experimental statistics. Statistics on the shadow pixels of the Ziyuan No. 3 remote sensing image data show that when T is set to 0.5, water bodies and shadow objects can be well distinguished.
步骤9：将步骤6中得到的大面积水体对象与步骤8中得到的小面积水体对象进行叠加，即将步骤6中得到的大面积水体对象和步骤8中得到的小面积水体对象求并集，得到卫星遥感影像的城市水体提取结果。Step 9: Superimpose the large-area water object obtained in step 6 with the small-area water object obtained in step 8, that is, the union of the large-area water object obtained in step 6 and the small-area water object obtained in step 8, The extraction results of urban water bodies from satellite remote sensing images are obtained.
一种卫星遥感影像的城市水体提取方法的总体流程图如图4所示。The overall flowchart of a method for extracting urban water bodies from satellite remote sensing images is shown in Figure 4.
为验证方法的有效性，分别采用NDWI方法水体提取结果和最大似然法(MaxLike)水体提取结果进行对比实验。在中国境内选取了5处不同地区且具有不同周边环境的影像用于实验，它们包括湖泊和河流，分别位于中国地区的北京市、武汉市、苏州市、广州市，其中武汉市选用了两个不同覆盖区域的影像。资源3号影像详细信息描述如表2所示，实验中详细的参数的设置情况如表3所示，其中band4的数值首先按下式归化到0-255取值范围，然后再选取阈值进行分割的。In order to verify the effectiveness of the method, the water body extraction results of the NDWI method and the water body extraction results of the maximum likelihood method (MaxLike) were used for comparative experiments. In China, five images with different surrounding environments were selected for the experiment. They include lakes and rivers, which are located in Beijing, Wuhan, Suzhou, and Guangzhou in China. Two of them were selected in Wuhan. Imagery of different coverage areas. The detailed information description of Resource No. 3 image is shown in Table 2, and the detailed parameter settings in the experiment are shown in Table 3, where the value of band4 is first normalized to the value range of 0-255 according to the formula, and then the threshold is selected for divided.
表2实验数据详细信息Table 2 Experimental data details
表3不同实验地区阈值设定(其中T1，T2，T3，分别为NNDWI1、NNDWI2、band4的分割阈值)Table 3 Threshold settings in different experimental regions (where T1, T2, T3 are the segmentation thresholds of NNDWI1, NNDWI2, and band4 respectively)
为便于不同方法分类结果的目视判读与分析，对正确分类的水体像元赋予灰色，正确分类的非水体像元赋予黑色，错误分类的像元赋予白色，实验结果如图5所示。从图5的分类结果目视判读可以发现，本发明提出的AUWEM的水体提取分类精度要好于NDWI的水体提取分类精度和最大似然法的水体提取分类精度。AUWEM对水体边缘混合像元可以很好的分类(结合Beijing、Wuhan_1和Wuhan_2地区的水体分类结果)、对细小的池塘水体检测性能好于NDWI和最大似然法(结合Suzhou地区水体分类结果)、对房屋建筑物的阴影可以很好的去除(结合Suzhou和Wuhan_2地区水体分类结果)。In order to facilitate the visual interpretation and analysis of the classification results of different methods, the correctly classified water body pixels are assigned gray, the correctly classified non-water body pixels are assigned black, and the misclassified pixels are assigned white. The experimental results are shown in Figure 5. From the visual interpretation of the classification results in Fig. 5, it can be found that the water extraction and classification accuracy of AUWEM proposed by the present invention is better than that of NDWI and the maximum likelihood method. AUWEM can classify the mixed pixels at the edge of the water body very well (combined with the water body classification results in Beijing, Wuhan_1 and Wuhan_2 areas), and the detection performance of small pond water bodies is better than NDWI and maximum likelihood method (combined with the water body classification results in Suzhou area), The shadows of houses and buildings can be removed very well (combined with the results of water body classification in Suzhou and Wuhan_2 areas).
不同实验区域三种方法水体提取分类精度比较统计结果如表4所示，从表4的统计结果中发现，AUWEM水体提取分类精度要高于NDWI和最大似然法。AUWEM在这5个实验区域的分类精度是最高，平均Kappa系数达93.0％；而NDWI的分类精度最低，平均Kappa系数约为86.2％；最大似然法分类精度介于两者之间，平均Kappa系数约为88.6％。Table 4 shows the comparative statistical results of the three methods of water body extraction and classification accuracy in different experimental areas. From the statistical results in Table 4, it is found that the classification accuracy of AUWEM water body extraction is higher than that of NDWI and maximum likelihood method. AUWEM has the highest classification accuracy in these five experimental areas, with an average Kappa coefficient of 93.0%; while NDWI has the lowest classification accuracy, with an average Kappa coefficient of about 86.2%; the classification accuracy of the maximum likelihood method is between the two, with an average Kappa coefficient of 93.0%. The coefficient is about 88.6%.
表4不同实验区域三种方法的精度统计Table 4 Accuracy statistics of the three methods in different experimental areas
为了更加详细的评估AUWEM对水体的提取精度，采用生产者精度、用户精度、Kappa系数、漏检率、虚警率、总错误率6个指标来描述方法的水体提取精度。三种不同方法在5个实验地区的6个指标直方图如图6(a)～(f)所示，从各个指标的直方图可以发现，AUWEM的水体提取分类精度要高于NDWI的水体提取分类精度和最大似然法的水体提取分类精度。AUWEM在水体提取的虚警率方面，除在Suzhou地区达到9.1％左右，其他的实验区域都是低于5％；在水体漏检率方面，5个地区漏检率都要明显少于NDWI和最大似然法。当水体提取的虚警率与水体漏检率都低时，总错误率自然也就是最低。从图6(a)～(f)中可以看出AUWEM总错误率最低，其次是最大似然法，总错误率最高的是NDWI，其中AUWEM平均总错误率约为11.9％，最大似然法平均总错误率约为18.2％，NDWI平均总错误率约为22.1％％。In order to evaluate the water body extraction accuracy of AUWEM in more detail, six indicators including producer accuracy, user accuracy, Kappa coefficient, missed detection rate, false alarm rate, and total error rate are used to describe the water body extraction accuracy of the method. The histograms of 6 indicators in 5 experimental areas with three different methods are shown in Figure 6(a)-(f). From the histograms of each indicator, it can be found that the classification accuracy of water body extraction by AUWEM is higher than that of water body extraction by NDWI Classification Accuracy and Classification Accuracy of Water Body Extraction by Maximum Likelihood Method. In terms of the false alarm rate of water body extraction by AUWEM, except for the Suzhou area, which reached about 9.1%, the other experimental areas were all lower than 5%. Maximum Likelihood Method. When the false alarm rate of water body extraction and the water body missed detection rate are both low, the total error rate is naturally the lowest. From Figure 6(a)~(f), it can be seen that AUWEM has the lowest total error rate, followed by the maximum likelihood method, and the highest total error rate is NDWI, in which the average total error rate of AUWEM is about 11.9%, and the maximum likelihood method The average total error rate is about 18.2%, and the NDWI average total error rate is about 22.1%.
对于水体提取分类生产者精度，AUWEM水体提取分类生产者精度最高，平均约为91.6％；最大似然法水体提取分类生产者精度次之，平均约为84.8％；NDWI的水体提取分类生产者精度最低，平均约为81.6％。在水体提取用户精度方面，最大似然法的水体提取用户精度要大于的AUWEM和NDWI，平均精度用户高达96.6％；AUWEM方法次之，平均精度用户约为96.4％；最差的仍然属于NDWI方法，平均精度用户约为95.7％。For the accuracy of water body extraction and classification, AUWEM has the highest accuracy of water extraction and classification, with an average of about 91.6%; the accuracy of maximum likelihood method for water extraction and classification is second, with an average of about 84.8%; NDWI’s accuracy of water extraction and classification is about 84.8%. The lowest, with an average of about 81.6%. In terms of water body extraction accuracy, the water body extraction accuracy of the maximum likelihood method is greater than that of AUWEM and NDWI, with an average accuracy of 96.6%; the AUWEM method is second, with an average accuracy of about 96.4%; the worst still belongs to the NDWI method , the average accuracy user is about 95.7%.
为了更加客观评价三种方法提取的水体的边缘检测精度，设计以下方法来进行精度评价，方法具体实施描述如下：In order to more objectively evaluate the edge detection accuracy of the water body extracted by the three methods, the following method is designed for accuracy evaluation. The specific implementation of the method is described as follows:
首先利用参考影像采用Canny算子获取三种方法提取的水体的边界线；Firstly, use the reference image and use the Canny operator to obtain the boundary line of the water body extracted by the three methods;
对获取的边界线进行数学形态学的膨胀处理，建立以边界线为中心半径为4个像元的缓冲区域；Carry out mathematical morphology expansion processing on the obtained boundary line, and establish a buffer area with the boundary line as the center and a radius of 4 pixels;
然后对缓冲区域的像元进行判断，假设缓冲区域总像元值为N，正确分类的像元数目的NR，漏检像元数目为No，虚警数目为Nc，则：Then judge the pixels in the buffer area, assuming that the total pixel value in the buffer area is N, the number of correctly classified pixels is NR , the number of missed pixels is No, and the number of false alarms is Nc, then:
其中，A+Eo+Ec＝100％。A表示边缘像元正确划分类别的比例，在这里称它为边缘像元正确分类精度；Eo表示边缘像元漏检的比例，称它为边缘像元漏检率；Ec表示边缘像元虚警的比例，称它为边缘像元虚警率。将经过数学形态学膨胀处理的水体的边界线与从资源3号卫星获取的遥感影像进行叠加，得到边缘精度评估中待评估区域。边缘精度评估中待评估区域获取示意图如图7所示。However, A+Eo+Ec=100%. A represents the proportion of edge pixels that are correctly divided into categories, which is called the correct classification accuracy of edge pixels; Eo represents the proportion of edge pixels that are missed, and is called the rate of edge pixels that are missed; Ec represents the false alarm rate of edge pixels It is called the edge pixel false alarm rate. The boundary line of the water body processed by mathematical morphology expansion is superimposed with the remote sensing image obtained from the Ziyuan-3 satellite to obtain the area to be evaluated in the edge accuracy evaluation. The schematic diagram of the area to be evaluated in the edge accuracy evaluation is shown in Figure 7.
根据上述方法统计了实验区域的水体边缘检测精度，分别统计了边缘像元虚警率(Commission Error)、边缘像元漏检率(Omission Error)和边缘像元正确分类精度(Accuracy of edge detection)，并对实验结果进行比较，如图8(a)～(c)所示。从图8的比较结果可以很清晰的发现：AUWEM方法边缘像元正确分类精度要好于NDWI和最大似然法。其中AUWEM方法水体边缘像元正确分类精度最高达93.7691％(Guangzhou地区)，最小精度为79.5798％(Wuhan_2地区)；NDWI水体边缘像元正确分类精度最高达84.0917％(Suzhou地区)，最小精度为69.8310％(Beijing地区)；最大似然法水体边缘像元正确分类精度最高达85.8149％(Guangzhou地区)，最小精度为69.7974％(Wuhan_2地区)。According to the above method, the water body edge detection accuracy in the experimental area was counted, and the edge pixel false alarm rate (Commission Error), edge pixel omission rate (Omission Error) and edge pixel correct classification accuracy (Accuracy of edge detection) were counted respectively. , and compare the experimental results, as shown in Figure 8(a)~(c). From the comparison results in Figure 8, it can be clearly found that the correct classification accuracy of edge pixels by AUWEM method is better than that of NDWI and maximum likelihood method. Among them, the correct classification accuracy of water edge pixels by AUWEM method is up to 93.7691% (Guangzhou area), and the minimum accuracy is 79.5798% (Wuhan_2 area); the correct classification accuracy of NDWI water edge pixels is up to 84.0917% (Suzhou area), and the minimum accuracy is 69.8310 % (Beijing area); the correct classification accuracy of water edge pixels by maximum likelihood method is up to 85.8149% (Guangzhou area), and the minimum accuracy is 69.7974% (Wuhan_2 area).
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611223281.2ACN106650812B (en) | 2016-12-27 | 2016-12-27 | A Method for Extracting Urban Water from Satellite Remote Sensing Images |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611223281.2ACN106650812B (en) | 2016-12-27 | 2016-12-27 | A Method for Extracting Urban Water from Satellite Remote Sensing Images |
| Publication Number | Publication Date |
|---|---|
| CN106650812A CN106650812A (en) | 2017-05-10 |
| CN106650812Btrue CN106650812B (en) | 2019-08-06 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611223281.2AExpired - Fee RelatedCN106650812B (en) | 2016-12-27 | 2016-12-27 | A Method for Extracting Urban Water from Satellite Remote Sensing Images |
| Country | Link |
|---|---|
| CN (1) | CN106650812B (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107167431B (en)* | 2017-05-26 | 2019-07-05 | 中国科学院遥感与数字地球研究所 | A method and system for identifying black and odorous water bodies based on spectral index model |
| CN107688776B (en)* | 2017-07-21 | 2022-12-16 | 同济大学 | A kind of urban water extraction method |
| CN107564017B (en)* | 2017-08-29 | 2020-01-10 | 南京信息工程大学 | Method for detecting and segmenting urban high-resolution remote sensing image shadow |
| CN107977968B (en)* | 2017-12-22 | 2021-03-19 | 长江勘测规划设计研究有限责任公司 | Building Hierarchical Detection Method Based on Building Shadow Information Mining |
| CN108830844B (en)* | 2018-06-11 | 2021-09-10 | 北华航天工业学院 | Facility vegetable extraction method based on multi-temporal high-resolution remote sensing image |
| CN109544558B (en)* | 2018-09-27 | 2021-05-11 | 浙江工业大学 | A method for separating shadows and water bodies in complex urban environments |
| CN109325973B (en)* | 2018-11-19 | 2020-11-24 | 珠江水利委员会珠江水利科学研究院 | An atmospheric correction method for water bodies in urban river network areas |
| CN110826394A (en)* | 2019-07-16 | 2020-02-21 | 北京大学 | Reservoir identification method and device based on convolutional neural network algorithm |
| CN110688923A (en)* | 2019-09-19 | 2020-01-14 | 中国电子科技集团公司第二十九研究所 | An urban waterlogging risk area extraction method based on Sentinel 1A SAR data |
| CN111275631A (en)* | 2020-01-08 | 2020-06-12 | 中国科学院东北地理与农业生态研究所 | Method for eliminating shadow interference during urban water body extraction by remote sensing image |
| CN113516084B (en)* | 2021-07-20 | 2023-04-25 | 海南长光卫星信息技术有限公司 | Semi-supervised classification method, device, equipment and medium for high-resolution remote sensing image |
| CN114359243B (en)* | 2022-01-10 | 2022-08-26 | 首都师范大学 | Seasonal small micro-wetland dynamic monitoring method |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102054274A (en)* | 2010-12-01 | 2011-05-11 | 南京大学 | Method for full automatic extraction of water remote sensing information in coastal zone |
| CN103400151A (en)* | 2013-08-16 | 2013-11-20 | 武汉大学 | Optical remote-sensing image, GIS automatic registration and water body extraction integrated method |
| CN103914692A (en)* | 2014-04-21 | 2014-07-09 | 山东科技大学 | Remote sensing recognition method for surface water systems around coal mine |
| CN105809140A (en)* | 2016-03-18 | 2016-07-27 | 华南农业大学 | Method and device for extracting surface water body information based on remote sensing model |
| CN106023133A (en)* | 2016-04-26 | 2016-10-12 | 武汉大学 | High resolution remote sensing image water body extraction method based on multi-feature combined treatment |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB0808340D0 (en)* | 2008-05-08 | 2008-06-18 | Univ Edinburgh | Remote sensing system |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102054274A (en)* | 2010-12-01 | 2011-05-11 | 南京大学 | Method for full automatic extraction of water remote sensing information in coastal zone |
| CN103400151A (en)* | 2013-08-16 | 2013-11-20 | 武汉大学 | Optical remote-sensing image, GIS automatic registration and water body extraction integrated method |
| CN103914692A (en)* | 2014-04-21 | 2014-07-09 | 山东科技大学 | Remote sensing recognition method for surface water systems around coal mine |
| CN105809140A (en)* | 2016-03-18 | 2016-07-27 | 华南农业大学 | Method and device for extracting surface water body information based on remote sensing model |
| CN106023133A (en)* | 2016-04-26 | 2016-10-12 | 武汉大学 | High resolution remote sensing image water body extraction method based on multi-feature combined treatment |
| Publication number | Publication date |
|---|---|
| CN106650812A (en) | 2017-05-10 |
| Publication | Publication Date | Title |
|---|---|---|
| CN106650812B (en) | A Method for Extracting Urban Water from Satellite Remote Sensing Images | |
| Shi et al. | Automatic road crack detection using random structured forests | |
| CN103400151B (en) | The optical remote sensing image of integration and GIS autoregistration and Clean water withdraw method | |
| CN102435713B (en) | Automatic detection system for quality of traditional Chinese medicine | |
| CN107038416B (en) | A Pedestrian Detection Method Based on Improved HOG Feature of Binary Image | |
| CN110097537B (en) | Meat quality quantitative analysis and evaluation method based on three-dimensional texture features | |
| CN106294705B (en) | A kind of batch remote sensing image preprocess method | |
| CN110705565A (en) | Lymph node tumor region identification method and device | |
| CN106096563A (en) | Plant automatic recognition system and method | |
| CN112837290A (en) | A Crack Image Automatic Recognition Method Based on Seed Filling Algorithm | |
| CN111199195B (en) | A method and device for fully automatic monitoring of pond status based on remote sensing images | |
| CN109376740A (en) | A kind of water gauge reading detection method based on video | |
| CN112765556B (en) | A new mangrove age assessment method based on dense temporal remote sensing data | |
| CN106971397B (en) | Based on the city high-resolution remote sensing image dividing method for improving JSEG algorithms | |
| CN109034269A (en) | A kind of bollworm female male imago method of discrimination based on computer vision technique | |
| CN104636755A (en) | Face beauty evaluation method based on deep learning | |
| CN111007013B (en) | Crop rotation fallow remote sensing monitoring method and device for northeast cold region | |
| CN116703932A (en) | CBAM-HRNet model wheat spike grain segmentation and counting method based on convolution attention mechanism | |
| CN110766708A (en) | Image comparison method based on contour similarity | |
| CN114965316A (en) | Hyperspectral multi-parameter based water quality monitoring method and computer program product | |
| CN106022303A (en) | Fresh water algae rough classification and counting method based on lensless holographic imaging | |
| CN107657246B (en) | A Building Detection Method Based on Multi-Scale Filtering Building Index in Remote Sensing Image | |
| CN118608977A (en) | A method for detecting changes in cultivated land for non-grain use based on remote sensing images | |
| CN106324708B (en) | Digitizing solution, the device of rainfall record drawing | |
| CN116343002A (en) | Image analysis sewage treatment sludge area sedimentation velocity prediction method |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | Granted publication date:20190806 Termination date:20211227 | |
| CF01 | Termination of patent right due to non-payment of annual fee |