基于合成孔径雷达的农作物后向散射特性及纹理信息分析——以吉林省农安县为例

Backscattering characteristics and texture information analysis of typical crops based on synthetic aperture radar: A case study of Nong'an County, Jilin Province

  • 摘要: 及时掌握农作物类型、时空分布和结构信息,是合理调整农业结构的重要科学依据。针对光学遥感依赖于太阳辐射,在农作物生长周期内常受制于云雾的影响而无法获取到光学遥感数据的问题,本文采用全天时全天候、不受云雾等天气影响的合成孔径雷达(synthetic aperture radar,SAR)影像,探讨典型农作物的后向散射特性和纹理特征,为采用合成孔径雷达影像实现高精度农作物大面积监测提供科学依据。以吉林省农安县为例,利用12景Sentinel-1B双极化SAR影像数据,经影像预处理和统计分析,研究不同极化的农作物后向散射特性和纹理信息。结果表明:3种作物(大豆、玉米、水稻1和水稻2)同极化(VV)的SAR后向散射系数在生长周期内均高于交叉极化(VH),农作物植株形态改变极化方式的能力为-25~-15 dB。3种作物在整个生长期内,后向散射系数呈现较大波动,各阶段后向散射特征差异明显。在生长初期,土壤对后向散射特征起主要作用,在SAR图像上表现为暗色调;随着作物生长,冠层散射叠加土壤散射作用占据主要位置,散射值随作物生长呈现逐渐增加的趋势,在SAR图像上表现为亮色调;拔节(分枝)后(7月10日后)作物的后向散射信号除冠层散射作用外,还主要受到土壤含水量及其与作物相互作用的影响,因此拔节后两种水稻后向散射系数下降幅度较大。水稻对雷达波的吸收强于玉米和大豆,整体上后向散射系数第2种水稻<第1种水稻<玉米<大豆,尤其在VH极化方式下表现更明显。对作物SAR纹理信息的研究表明纹理信息的均值、方差和协同性对于农作物的SAR识别更有效,最佳纹理信息为VH极化均值,有利于识别3种作物;VV极化方差和VV协同性有助于区分两种水稻;SAR影像识别作物的最佳时相为5月23日至7月10日。

     

    Abstract: Agriculture is the foundation of the national economy. Clarifying crop type, spatial and temporal distribution, and planting structure is an important scientific basis for the rational adjustment of agricultural framework. Optical remote sensing relies on solar radiation and is often subject to the influence of clouds. Therefore, optical remote sensing images cannot be obtained during cloudy weather. Synthetic aperture radar (SAR) images are not affected by cloud and fog and can provide all-day, all-weather data. Therefore, the backscatter characteristics and texture information of typical crops can be analyzed by using SAR images, which may improve accuracy of large area monitoring of crops in all weather conditions. In this study, Nong'an County in Jilin Province was taken as the study area, and 12 Sentinel-1B SAR images with dual-polarization were pre-processed. Then the SAR backscattering characteristics and texture information of typical crops with dual polarization were analyzed. The results showed that the co-polarization (VV) SAR backscattering coefficients of three crops (soybean, corn, rice 1 and rice 2) were higher than the cross-polarization (VH) coefficients during the growth period, and the ability of crop plants to change the polarization mode was about -25 to -15 dB. Over the whole growth seasons of crops, the backscattering coefficients showed large fluctuations. The backscattering characteristics of each growth stage were different. During the early growth stage of three crops, the soil played a major role in determining the backscattering characteristics, and the roughness of soil was small. Therefore, the three crops on the SAR images were dark in color. With the growth of the crops, the backscattering from crops was mainly due to canopy scattering and soil scattering. The backscatter coefficient value increased with crop growth, which meant that the color of the crops changed to a relatively bright shade on the SAR image. When crops reached the jointing stage or branching stage (after July 10), in addition to canopy scattering, their backscattering signal was strongly affected by the soil water content and its interaction with respective crops. The soil water content of rice was high. Therefore, the backscattering coefficient of rice considerably decreased after the jointing stage. The absorption of radar waves by rice was stronger than those of corn and soybean, which meant that the backscattering coefficient of rice was smaller than the coefficients for corn and soybean, especially in the VH polarization SAR images. The crop SAR texture information analysis showed that the mean, variance, and homogeneity could accurately identify crops on SAR images. The best texture information was the mean of VH polarization. The variance and homogeneity of VV polarization could also accurately identify the two kinds of rice. The best recognition phase for the crops on SAR images was from May 23 to July 10.

     

/

返回文章
返回