Automatic Recognition Of Typical Natural Elements From GF-2 Remote Sensing Imagery

The remote sensing science and technology started a little later,but after decades of continuous development,remote sensing technology has become an important means of obtaining spatial information,and has been widely used in surveying and mapping,geoscience and other fields.With the development of high-spatial and high-resolution remote sensing technology,a sky-space-ground integrated remote sensing system has been formed,and it is continuously being penetrated into all sectors of society.Remote sensing images are generally large in size and have complex backgrounds.Contains many ground objects with different sizes,different attitudes,and different directions.For a long time,the interpretation of remote sensing images has mainly relied on manual labor.Human interpretation generally requires prior knowledge,and the results are highly accurate,but it is time consuming,laborious,and costly.With the continuous improvement of remote sensing image resolution,the amount of remote sensing data has increased significantly,and the demand for geospatial information has increased significantly.Faced with a large amount of data and increasing market demand,manual interpretation methods are slightly stretched,but the automatic interpretation of remote sensing images is still immature and has not reached the level of practicality.Therefore,indepth study of remote sensing image automation and intelligent interpretation is a research topic with practical significance.The research content and conclusion of this article are as follows:1.This paper proposes an automatic cloud detection method for domestic high-resolution remote sensing satellite images based on multi-channel information.This method first improves the RGB to HIS color space of the remote sensing image,uses the saturation information and intensity information of the remote sensing image to generate the base map,and uses the hue information and near-infrared information to optimize the base map to generate a correction map.Then use Otsu threshold segmentation and histogram equalization segmentation to extract texture information and generate cloud seed map.Finally,the cloud information is extracted from the intensity information of the remote sensing image combined with the cloud seed map.The efficiency of cloud extraction of remote sensing images is greatly improved,and it is easy to use and adaptable.2.This paper proposes a water body automatic extraction model that depends on the characteristics of the image itself.The model takes the inherent characteristics of the water body on the remote sensing image as the main recognition features,and introduces the visual recognition results as recognition samples into the model.Only the red band,green band,and blue band are required as the calculation data,and no auxiliary information is required in other bands.The minimum distance and automatic threshold calculation methods are proposed,which effectively retain the characteristics of small samples.The experimental results prove that the water recognition result output by the model is consistent with the human visual interpretation result,and has the characteristics of strong adaptability.3.This paper proposes an automatic vegetation extraction method based on visual features.First,the remote sensing image is converted from RGB to HIS color space without the need for auxiliary information in the near-infrared band.Then,the hue characteristics of the vegetation are used to perform coarse extraction of the vegetation,and the coarsely extracted remote sensing image and the saturation image and intensity image are Gaussian transformed.Finally,the result of Gaussian transformation is input into the PM model for decision.The experimental results show that the vegetation results identified by this model are basically consistent with the visual recognition results,and have strong practicability.