| As the continuous enrichment of data acquisition techniques, the number of satellites on orbit has increased dramatically at home and abroad, and the satellite remote sensing platform has been rapidly developed. As early as institutions as NASA established in America, satellite played a significant role as an important method of earth observation. However, due to the limitations of imaging environment, revisiting period, payload and other factors, a single satellite cannot obtain data of multi-class, multi-layered, and multi-dimensional in a large range, and the information it provides has been unable to completely meet the urgent need of wide applications such as resource, environment, society, disaster. Therefore, space-borne observation system has gradually developed into a new phase of multi-sensor, multi-spectral and multi-temporal. Radar data and optical image are the main dataset of satellite remote sensing platform. Due to the current bound of techniques and limitations of traditional thinking, joint matching technology for multi-source, multi-spectral and multi-temporal satellite imagery is not yet mature, and there are still many problems to be solved. Therefore, it has great significance to explore new methods of image matching and have a further research of joint processing method for multi-source, multi-spectral and multi-temporal satellite imagery.This thesis innovatively discussed a registration method to process different sensor and temporal SAR and optical images. The dense matching of optical stereo images produced DEM, which can improve the data precision of DEM extracted from InSAR. And based on the two kind of data, two typical remote sensing data application have been done, which involves key technologies including data registration of different source, phase unwrapping method for InSAR, DEM fusion method of different source, cloud detection method of optical image and crop cover interpretation of optical image. The main work is as follows:1. Data registration of different source. Due to the significant direct georeferencing divergence of the attitude parameters of satellite images, the mainstream registration method cannot be suitable. Through analyzing the problems of the existing method and considering the characteristics of multi-source and multi-temporal, an overall-to-local and coarse-to-fine registration method (RoPS, Registration Based on Phase Correlation and Spline Function) has been designed. The method involves technologies including predetermination model of initial value of matching, accurate registration model based on spline functions with local control, gross error elimination. It has two main characteristics: firstly, it combined the matching strategy based on general model and statistics method based on spline functions with local control during the progress of registration; secondly, when the registration model is being calculated, gross error elimination of high coupling has been avoided, instead, the self-organizing point location sifting method has been added for the detection and elimination of gross error. The registration experiment has proved that the registration method of different method in this thesis can obtain high accuracy result of the registration between SAR and optical data.2. DEM quality assessment and integration. Long-term's practices of space borne InSAR have been proved that DEM has strong continuity but performs badly at detail when observe the land in large range, especially in the mountainous area whose landform undulated terribly. Accuracy of DEM is susceptible due to geometrical distortions including layerover, foreshortening and shadow. With the support of stereo images of Resources satellite three (ZY-3), accuracy and reliability of DEM can be improved by fusion algorithm. Generally speaking, taking DEM as the redundant observation information can effectively estimate the baseline and quickly processed the data, thus the DEM will be extracted. However, classic algorithm may lead to unpleasant result when affected by the area where result of phase unwrapping is bad. Inspired from the single image iteration mapping of optical image, the InSAR interferogram is taken as an equivalent central projection tomography. Based on the reference of SRTM(Shuttle Radar Topography Mission), the geometry plane coordinate of phase point has been iterated to detect bad unwrapping area in interferogram, and offer quality evaluation for the follow up elevation extraction. Furthermore, it can indicate the fusion of multi-source and improve the accuracy and veracity of DEM.3. Application of multi-source remote sensing data. In this paper, two typical application has been discussed from the fusion application of space borne InSAR and optical image:(1) Cloud detection of optical image. The cloud influenced the process of optical images in many ways, such as image matching, image fusion, image mosaic. There are two main problems:firstly, cloud and other highlight such as snow or city are hard to distinguish; secondly, the data quantity is mass and it is hard to achieve good effect without artificial intervention. The radar beam can penetrate the cloud. According to this, SAR images in the same region are used to assist the cloud detection.(2) Crop cover interpretation. The geometry and radiation information of single sensor image cannot be both taken into account. Therefore, joint observation of different sensor is inevitable in order to obtain information of high spatial resolution, high temporal resolution and high spectrum resolution. Combining the optical images of high spatial and high temporal resolution and DEM, a detection algorithm to interpret the crop cover has been designed in this paper. The method can be applied in crop estimation and monitoring of key kinds of crop.Through the idea of fusion of space borne SAR and optical data, and referred from the open geographic information data, namely, SRTM, a registration algorithm between SAR and optical image has been designed in this paper. Using the DEM data obtained from optical images, the DEM extracted from InSAR has been improved. This thesis explored two fusion application and implemented the technology strategy of related algorithm. The experimental results using real data proved that the algorithm proposed by this paper can be well used. Therefore, this algorithm lays a foundation to the combined earth observation processing with multi-source, multi-temporal satellite data in the future.To be mentioned, the method proposed in this paper has been successfully applied in several projects. The accuracy and veracity of matching have been proved in projects involving series of Chinese resource satellites and GF-1. |
PDF Full Text Request