Segmentation Of High Resolution Remotely Sensed Image Of Earthquake Disasters Based On Intelligent Optimization Algorithm

China is one of the countries with serious earthquake disaster in the world. Rapid access to the information of disaster is the key to emergency rescue. In traditional method, artificial scene investigation is used to obtain information about the earthquake disaster. However, this traditional method consumes large amounts of manpower, material resources, and time, which makes it unable to meet the requirements of emergency rescue of earthquake. Due to the characteristics of macro and real-time, remote sensing image has become one of the major means of information extraction in seismic images. In the extraction process, image segmentation is the key step, and is becoming a hotspot and difficult problem.This paper analyzed the current difficulties faced by the seismic image segmentation first, and then introduced some intelligent algorithms including gravitational search algorithm(GSA). Afterwards, two kinds of improved GSA were used to high resolution remote sensing images, The segmentation of high resolution remotely sensed seismic images was carried out based on two kind of improved GSA. The main contents of this thesis are shown as follows:1. Study on feature sets extraction of typical targets in the seismic disaster images. The feature sets are extracted from different domains, including space domain and frequency domain. The extracted features include spectrum features as well as texture features of typical targets in the seismic disaster images.2. Study on combination and optimization of high dimensional features. GA-GSA algorithm was proposed based on in-depth analysis of GA(genetic algorithm) and GSA. And, binary coding GA-GSA was used to carried out combination and optimization of feature sets obtained before.3. Study on feature clustering and image segmentation. The memory characteristic of particle swarm algorithm(PSO) was introduced to improve the performance of GSA(called MGSA). Then, a hybrid algorithm of K-means and MGSA is used to overcome the disadvantages of K-means including low efficiency and premature convergence. Finally, the high resolution remotely sensed seismic images was segmented successfully with high precision and efficiency.