Research Of Crop Recognition Method Based On Multi-temporal C-band Quad-Poi SAR

Satellite remote sensing, providing all kinds of crops in spatial-temporal distribution information, has an advantage over traditional surface investigation methods. It is not only macroscopic, dynamic, timely and precise, and it is also cheaper and less time-consuming. And now it has gradually become one of the important data sources and main methods in the aspect of crop monitoring. Because of humid climate and fragmental distribution of the ground objects in the regions of south China and southwest of China, a variety of crops mixing planting is a common planting pattern, which causes a big difference in phenolgical characteristics between the different types of crops and the different planting stages of the same crop. It is hard to effectively recognize and monitor the land cover simply relying on the general optical remote sensing data. However, with the all-weather and day-night time imaging capability, and without being restricted by extreme weather conditions, Synthetic Aperture Radar(SAR) plays a key role in detecting the vast areas in southern parts of China, especially in Guangdong Province. But in the previous research of agricultural remote sensing based on the polarization SAR data,the data polarization mode is single, which cannot show scattering characteristics of crops fully. The long time series data also has been taken into consideration rarely. Therefore, not only can it not display the crops growth regularity and phenological characteristics correctly, but it even makes lower identification accuracy.Therefore, this paper research on the crop recognition method based on multi-temporal C-Band Quad-Pol SAR, taking Zhanjiang of Guangdong province as a case study and taking the multi-temporal C-Band Quad-Pol SAR as the data resources. Based on all kinds of polarization target decomposition models and other parameters features, the thesis studies C-Band polarization characteristics of the typical crops in the experimental area, analyzes the growth features of the typical crops such as sugarcane, rice, banana in a period and establishes the temporal variation curves of the polarization characteristics, the discrete degree of the feature space and space power distribution model on the foundation of the polarization synthesis. According to the thinking above, the study continues exploring the crop identification method, not only considering the differences of polarization characteristics between different types of crops, but also thinking about the diversity of planting times and date of the same crop. Comparing with the differences of optimal classification feature combination and phase combination of typical crops in the study area, the research establishes the recognition knowledge and rules of the typical crops and applies them based on object-oriented classification. The results indicate that, choosing the image on Mar. 28 th as the optimal phase, the multi-polar feature combination which is made up of the polarization scattering angles α, polarization scattering entropy H, volume scattering intensity PVOL, shows excellent classification performance on the temporal variation curves of the polarization characteristics and the discrete degree of the feature space, and presents more obvious differences of the polarization characteristics on the space power distribution of the typical crops such as sugarcane, rice and banana. Through constructing indexes and threshold segmentation method to establish the object-oriented classification rules to extract the distribution information of the sugarcane, banana and rice, the final classification accuracy is above ninety-two percent and the kappa coefficient is 0.86, which suggests that the method effectively improves the classification precision of crops and enhances the identification accuracy as well. To sum up, the classification characteristics of the C-Band Quad-Pol SAR have a great advantage on identifying the typical crop such as sugarcane, banana and rice. The research not only has certain scientific significance on the crop identification in southern China, also has important practical meaning in promoting the multi-temporal full polarization SAR data in the more extensive application of crop condition monitoring and crop yield estimation and so on.