| Maize is an important food, forage and industrial crop in China. The monitoring of nationwide maize planting area and growing status needs an objective, precise and reliable technical support system as a guarantee. With the development of remote sensing technology, it is possible to monitor the large acreage of crops and their growth status, which provides an important reference for the policy of national agricultural production and food trade. But China has complicated geography and climate types in the eastern region, where the climate condition is cloudy and rainy which limits the monitoring by optical remote sensing of maize monitoring; it is difficult to obtain the full image data during the growth period of maize. The radar remote sensing becomes the other main source of maize monitoring data which have advantage in the all-time and all-weather imaging capability.In this paper, using the multi-temporal full-polarization Radarsat-2 SAR data of growth period of maize and the growth data of ground crop, taking Suining county, Jiangsu province as the study area, maize mapping principles and methods were analyzed and discussed, then compared the method of decision tree classification and the method based on scattering mechanism, these two classification methods were tested with the measured ground GPS data; On this basis, the relationships between the backscattering coefficient of maize and its biological parameters were analyzed. And the biological parameters of maize were retrieved by the empirical model, verified its accuracy according to the ground measured data. The result of this paper show that monitoring of maize planting area and growing status were feasibility and accuracy by using the Radarsat-2 SAR data, and its provide an important reference for extracting, identifying and testing crop information from radar remote sensing image. Hence, the main conclusions are as follow:(1) Pre-processing of Radarsat-2 SAR data. The calibration, speckle noise filtering, and geo-reference are finished by using the software such as SARINFOS, PLOSARPRO, ENVI and ERDAS. The co-registration is completed between three radar images and the ground GPS data, which is providing the standard data source for crop information extraction.(2) Research of maize identification and mapping method. According to the backscattering of different landsurface objects characteristics (building, water, maize, rice, forest, and soybean) of multi-temporal full-polarization Radarsat-2 SAR data, the most different backscattering characteristics of different crop of SAR data (July 30, 2009) were selected. Maize mapping were researched by using decision tree method and scattering mechanism method. The results show that both methods can extract the maize distribution information from the SAR image in the similar accuracy. However, due to SAR image suffers from speckle noise, the identification of decision tree classification was seriously affect. More smooth maize mapping and higher maize mapping accuracy can be reached by the mapping classification of scattering mechanism based on Wishart supervised method.(3) Methods of mapping results validation. Using the confusion matrix and pixel statistics of maize classification, the results of two classification methods are verified combined with ground GPS data. Comparing the results of verification methods, pixel statistics methods are more convenient effective and reliable than confusion matrix method, according to the same training samples.(4) Using the correlation of Radarsat-2 SAR data and ground maize growth parameters, empirical models are established to retrieved the main growth parameters (LAI) and monitor the growing of maize according to the results of maize mapping. The results show that the HH, HV polarization backscattering coefficient in SAR images have a good correlation with the leaf area index of maize, and the growing conditions of maize can be inverted with the distribution image of maize planting area.
|
PDF Full Text Request