|
Remote sensing image classification is an important aspect of the remote sensing image processing, and is the basis for subsequent extraction of thematic information to detect dynamic changes in the production of thematic maps, build remote sensing database. Widely used in various fields along with the development of the modern remote sensing technology, and remote sensing technology, remote sensing image data along with the rapid expansion, have become more sophisticated remote sensing image classification accuracy and real-time processing requirements. Now, research has focused on improving the efficiency of the classification process of the new algorithm proposed can not significantly improve the accuracy and timeliness of the sorting process, however, can not meet the needs of practical application. The papers in the study of existing classification algorithm based on analysis of the classification process steps affecting the classification accuracy and speed of rapid classification level of the overall process. The main idea is: the steps for sample collection, the design is based on AOI (Area of ??Interest region of interest)) the classified sample database and management system, to improve the efficiency and accuracy of sample collection; classification calculation steps to improve the classification algorithm parallelization processing speed. The main research work as follows: First, the analysis of several commonly used classification algorithm, summarized the general process of the classification process, and found that the impact speed of the sorting process two major steps sample selection and classification. Former speed is limited by the operator level and aids choice, and the accuracy of the direct impact on the accuracy of the classification results; latter speed is mainly affected by the impact of the algorithm and hardware, and thus determine the main issues to be addressed in this article. Secondly, these two issues are looking for solutions. The calculation procedure for the classification, designed and implemented a common classification algorithm parallelization. The algorithm uses a master-slave mode, the use of data-based parallelization strategies, divided image, copy parameters, classification processing speed has been greatly improved. For the sample selection step, designed and implemented based on the AOI the classified sample database and its management system, thereby reducing the operating personnel professional level of classification results, while improving the efficiency of sample selection. Finally, on the basis of the algorithm and database, the use of engineered, modular, process-oriented, hierarchical design mode, client, server-based architecture, designed a set of remote sensing image parallel classification system, the use of the system can be achieved The bulk of remote sensing images is fast, convenient classification process. And parallel classification experiments on the system to verify the efficiency of parallel sorting process and discuss the parallel classification calculated speedup factors.
|