Research On Retrieval And Recognition System Based On Mountain Area Remote Sensing Image

With the rapid development of remote sensing technology,the amount of high spatial resolution and hyperspectral remote sensing image data obtained by ground stations has increased dramatically.From the image database with a large number of remote sensing images,efficiently and accurately obtaining the user interested image information and effectively recognizing the target from the remote sensing image are both domestic and foreign research hotspots and have a high research value.Taking the mountainous area remote sensing images of GF-1 as the research object,a method of hierarchical extracting the effective feature components is proposed and employed to improve the retrieval efficiency on remote sensing image classification retrieval.Based on the combination of color features and Hough transform,the effective recognition on roads is realized.And,under the.NET Framework,the interface of the remote sensing image retrieval and road recognition systems is designed by C#.The main works in this paper are described as follows:(1)GF-1 remote sensing image database of typical villages,ridges and terraces in Yangyuan,Hebei,is made.And,a method of hierarchical extracting the effective feature components is proposed.First,multiple features are extracted from the typical images with three different categories.Then,the feature of remote sensing image is extracted by stratification: the first layer,by analyzing the contribution value of all characteristic components in typical training remote sensing image samples,the effective feature components are extracted.The other layers,that is,the contribution value of the feature components selected in the upper layer of the remote sensing image is trained by the current layer,and then the effective feature of the current layer is analyzed until that the least effective feature is selected.The efficient feature component selection would be apply to remote sensing image classification and retrieval.(2)Combining the color characteristics of freeway with Hough transform is used to identify the roads from mountain area remote sensing images.According to the small range of road color threshold,the background in the part of road is removed.Then,based on extracting the edge,the image denoising by filter,two value of the image,the roads in the images are identified by using Hough transform line detection and circle detection.(3)The visual system of remote sensing image retrieval and recognition system is designed,including the visual main interface of the program,the visual system of the remote sensing image retrieval and the visual system of the road recognition in remote sensing image.The programs of visual system are written by C# and based on the.NET Framework.In order to realize the visual function of each module in the visual interface,the simulation programs of MATLAB need to be converted into the.dll format,and the data exchange between the MATLAB programs and the C# programs through the intermediate type of the MWArray.In the C# programs,the MWArray.dll file and the.dll format converted from MATLAB simulation program are referenced.The visual interface realizes the experimental results display of the remote sensing image retrieval and image road recognition.