结合改进的降斑各向异性扩散和最大类间方差的SAR图像水体提取
作者简介:李玉(1963-),男,吉林长春人,博士,教授,主要从事遥感数据处理理论与应用基础研究。E-mail: liyu@lntu.edu.cn
收稿日期: 2018-10-18
要求修回日期: 2019-03-01
网络出版日期: 2019-06-15
基金资助
国家自然科学基金项目(41301479)
辽宁省自然科学基金项目(2015020090)
Waterbody Extraction from SAR Imagery based on Improved Speckle Reducing Anisotropic Diffusion and Maximum Between-Cluster Variance
Received date: 2018-10-18
Request revised date: 2019-03-01
Online published: 2019-06-15
Supported by
National Natural Science Foundation of China, No.41301479
Liaoning Province Natural Science Foundation, No.2015020090
Copyright
利用遥感成像技术获取地面水体信息对水资源调查、自然灾害评估、流域规划和生态环境监测等具有重要意义,其中SAR成像作为大范围地面监测的可靠数据源,拥有全天时、全天候、广覆盖等光学遥感系统所不具有的优点,在水体提取中得到了广泛的应用。但由于受SAR图像相干斑噪声的影响,现有水体提取方法难以迅速、精确提取SAR图像中复杂精细的自然水体结构。为此,提出一种结合改进的降斑各向异性扩散和最大类间方差的SAR图像水体提取方法。首先,利用降斑各向异性扩散滤波SAR图像,在迭代滤波过程中通过计算图像间平均结构相似度自适应控制迭代过程,使其同时保持精细边缘和纹理结构;然后,以类间方差最大为准则,自适应确定阈值,实现滤波结果图像二值化分割。在二值化分割结果中,搜索具有相同像元值且位置相邻的前景像元点组成的连通区域,使每个单独的连通区域形成一个被标识的块,通过获取这些块的几何参数来消除图像的误分割,精确划定真实的水体区域,以实现SAR图像水体提取。为了验证提出方法的准确性,将本文方法提取的水体边界与人工绘制的水体边界叠加,结果表明二者可较好吻合。同时,从视觉、提取精度和运行时间对本文方法与目前常用3种SAR图像水体提取算法的结果进行比较分析,其中本文方法的运行时间满足实时应用的要求,提取结果的边界在2个像元评级区重叠度均达到80%,明显优于其他方法且本文方法提取结果在边界及细节信息等视觉方面也更加显著。对结果的定性及定量评价表明本文方法的优越性。
李玉 , 杨蕴 , 赵泉华 . 结合改进的降斑各向异性扩散和最大类间方差的SAR图像水体提取[J]. 地球信息科学学报, 2019 , 21(6) : 907 -917 . DOI: 10.12082/dqxxkx.2019.180525
The use of remote sensing technology to obtain surface waterbody information is of great significance for water resource investigation, natural disaster assessment, watershed planning, and ecological environment monitoring. As a reliable data source for large-scale ground monitoring, SAR imaging has unique advantages that optical remote sensing systems of all-weather, all-weather, and wide coverage do not have, and has been widely used in waterbody mapping. However, due to the influence of speckle noise of SAR imagery, existing methods for waterbody mapping are difficult to extract the complex and fine natural water structures from SAR imagery quickly and accurately. In this paper, a new waterbody extraction method for SAR imagery based on improved speckle reducing anisotropic diffusion and maximum between-cluster variance was proposed. First, the SAR imagery were filtered by improved speckle reducing anisotropic diffusion. The iterative process was adaptively controlled by calculating the average structural similarity between imagery in the iterative filtering process, so that the fine edges and texture structure could be preserved simultaneously. Then, based on the criterion of maximum variance between classes, the threshold value was determined adaptively, and the binary segmentation of the filtered image was conducted. In the binarized segmentation result, connected foreground regions composed of the pixel points that have same intensity values and adjoin to each other spatially, are searched. In so doing, each connected region formed an identified block. By obtaining geometric parameters of these blocks, the false segmentation of the imagery was eliminated, and real waterbody areas were precisely identified based on the SAR imagery. To verify the accuracy of the proposed method, water boundaries extracted by this method were on manually drawn waterbody boundaries. Results of comparing the two methods show that they are pretty consistent with each other. Meanwhile, the results of our proposed method were compared with the results of three other kinds of water extraction algorithms commonly used for SAR imagery, in terms of the visual level, extraction accuracy, and running time. The running time of the proposed method meets the requirement of real-time application. The overlap degree of the extracted boundary in two pixels rating areas has reached 80%, which is obviously superior to other methods and the extraction results of the proposed method are also more significant in visual aspects such as boundary and detail information. The qualitative and quantitative evaluation shows the superiority of our proposed method.
Fig. 1 Original intensity maps of the SAR imagery图1 原始SAR强度图像 |
Fig. 2 Flowchart of the method of this paper图2 本文方法的流程 |
Fig. 3 SRAD edge detection diagram图3 SRAD边缘检测示意 |
Fig. 4 SRAD first-order neighborhood Diffusion diagram图4 SRAD一阶邻域扩散示意 |
Fig. 5 Flowchart of the pseudo waterbody removal algorithm based on connected-domain calibration图5 基于连通域标定的伪水体去除算法流程 |
Fig. 6 SARD filtering results图6 SARD滤波结果 |
Fig. 7 Extraction results and qualitative evaluation of this paper图7 本文方法提取结果和定性评价 |
Fig. 8 Pixel buffer boundary overlay analysis图8 像元缓冲区边界叠加分析示意 |
Tab. 1 Extraction accuracy of the 5 SAR images in the case of different radius rating zones (%)表1 在不同半径评级区的情况下5幅SAR图像的提取精度对比 |
重叠 | 1个像元 | 2个像元 | 3个像元 | 4个像元 | 5个像元 | |
---|---|---|---|---|---|---|
图1(a) | 21.05 | 68.90 | 90.21 | 96.76 | 99.23 | 99.54 |
图1(b) | 19.13 | 60.11 | 88.56 | 93.62 | 96.23 | 97.04 |
图1(c) | 18.39 | 60.29 | 82.16 | 90.32 | 92.90 | 94.12 |
图1(d) | 8.27 | 59.84 | 81.66 | 92.48 | 93.77 | 95.93 |
图1(e) | 20.29 | 69.14 | 91.01 | 92.81 | 95.60 | 97.05 |
Fig. 9 Comparison of extraction results of Otsu, level set and reference [17] method图9 Otsu、水平集和文献[17]方法提取结果比较 |
Fig. 10 Otsu, level set and literature [17] method contour overlay results图10 Otsu、水平集和文献[17]方法轮廓线叠加结果 |
Tab. 2 Extraction accuracy of the 5 SAR images with different radius rating zones by the compared algorithms (%)表2 对比算法在不同半径评级区的情况下5幅SAR图像的提取精度对比 |
图像 | 方法 | 重叠 | 1个像元 | 2个像元 | 3个像元 | 4个像元 | 5个像元 |
---|---|---|---|---|---|---|---|
图1(a) | Otsu | 3.88 | 12.57 | 18.82 | 23.60 | 28.04 | 31.71 |
水平集 | 11.42 | 37.24 | 53.89 | 62.94 | 68.86 | 72.36 | |
文献[17] | 12.58 | 42.52 | 59.64 | 64.87 | 67.45 | 69.31 | |
图1(b) | Otsu | 4.16 | 13.25 | 19.72 | 23.87 | 27.11 | 30.06 |
水平集 | 9.84 | 30.29 | 43.03 | 49.27 | 53.51 | 56.75 | |
文献[17] | 15.44 | 46.78 | 62.08 | 66.28 | 67.63 | 69.36 | |
图1(c) | Otsu | 2.79 | 8.72 | 12.79 | 16.52 | 17.86 | 19.37 |
水平集 | 4.85 | 10.92 | 17.00 | 21.56 | 22.71 | 23.31 | |
文献[17] | 9.27 | 26.79 | 34.08 | 37.23 | 38.48 | 39.13 | |
图1(d) | Otsu | 1.19 | 5.65 | 9.43 | 11.12 | 12.50 | 13.37 |
水平集 | 4.96 | 11.47 | 18.30 | 22.57 | 24.96 | 26.06 | |
文献[17] | 10.96 | 23.47 | 31.30 | 36.57 | 39.96 | 41.06 | |
图1(e) | Otsu | 5.75 | 18.30 | 27.91 | 34.50 | 39.36 | 43.48 |
水平集 | 7.71 | 24.89 | 37.83 | 46.21 | 51.96 | 56.40 | |
文献[17] | 15.41 | 49.18 | 69.24 | 76.46 | 78.39 | 79.30 |
Tab. 3 Comparison of extraction time between the compared algorithms and the method of this paper (s)表3 对比算法与本文方法的水体提取时间对比 |
方法 | 图1(a) | 图1(b) | 图1(c) | 图1(d) | 图1(e) |
---|---|---|---|---|---|
Otsu | 1.205 | 1.227 | 4.380 | 15.608 | 4.633 |
水平集 | 224.873 | 216.374 | 1007.716 | 5068.251 | 1179.581 |
文献[17] | 9.527 | 8.371 | 56.153 | 223.176 | 78.113 |
本文方法 | 4.382 | 4.178 | 15.322 | 29.644 | 16.013 |
The authors have declared that no competing interests exist.
[1] |
|
[2] |
[
|
[3] |
[
|
[4] |
[
|
[5] |
[
|
[6] |
[
|
[7] |
|
[8] |
|
[9] |
[
|
[10] |
[
|
[11] |
[
|
[12] |
[
|
[13] |
[
|
[14] |
|
[15] |
[
|
[16] |
[
|
[17] |
[
|
[18] |
[
|
[19] |
|
[20] |
|
[21] |
|
[22] |
|
[23] |
http://rs.ceode.ac.cn/satelliteintroduce/introRadarsats.jsp# RADARSAT-1
|
[24] |
http://cloud.satimage.cn/data_intro
|
[25] |
http://www.cresda.com/CN/sjfw/zxsj/index.shtml
|
[26] |
|
[27] |
|
[28] |
|
/
〈 | 〉 |