Study On Remote Sensing Image Fusion Method Based On Integrated Mean - Drift Segmentation

Sensor is a core component of the remote sensing information detection. Is to detect, collect and record information tools feature of electromagnetic radiation. With the innovation of science and technology, sensor design technology has been continuously improved, and now the endless variety of high-precision sensors for the detection of terrestrial remote sensing information to provide a more scientific and efficient techniques. Due to the same principle of the sensor involved, various sensors acquired image data ground information exists a big difference compared with the single sensor acquisition, multi-source data having different spatial, temporal, spectral resolution, can be detected more fully material information. He is a comprehensive multi-source remote sensing information in the same study area or the same goal, can reveal the nature of images from a variety of targets multifaceted, providing complementary physical properties obtained for a complete description of the target.Image fusion is usually divided into pixel level fusion, feature level fusion and decision fusion. Performing image fusion process to select the data source is one of the key technologies, image data can be the same combination of sensors and different sensors, different sensor data and can be combined according to the resolution of differences, the various combinations of data obtained fusion results vary. In this paper, the main choice of three different combinations of experimental data, the first group:QuickBird panchromatic sensor with high geometric accuracy multispectral data fusion; the second group:SPOT6 moderate geometric precision sensor panchromatic and multi-spectral data fusion; s three groups:QuickBird sensor and multispectral data and cross combination SPOT6 panchromatic sensor.In addition, the fusion algorithm selection is key image fusion technology, research focus of this article is from the perspective of the image area is divided to explore the pixel level fusion techniques of remote sensing images. Proposed based on remote sensing image fusion method mean shift segmentation, which is a fusion of Brovey improvements. Use this method to participate in the integration of panchromatic, multispectral data segmentation, segmentation results corresponding. On this basis, associated with different criteria for segmentation results correlate an associate guidelines:A full-color image segmentation results for related benchmarks; two related criteria:segmentation based on multi-spectral data correlation results; associate three criteria:the PAN and multispectral segmentation results superimposed association. The results obtained in three different correlation diagram, and then shipped to the associated results using Brovey fusion method. Finally, the use of ENVI Brovey integration tools original image fusion, fusion based on region segmentation results were analyzed. By visual interpretation and objective metrics to calculate the fusion of four qualitative and quantitative comparison of the results of the analysis, based on the fusion results in the spectral region segmentation retention and spatial resolution better than inheritance Brovey fusion results.