Study On Main Features Information Extraction Technology Of High-resolution Remotely Sensed Image Based On Multiresolution Segmentation

With the rapid development of sensor technology and remotely sensed platform,the technology of remote sensing observation has gradually matured. High spatialresolution is the general developing trend of remote sensing. In the low-moderateresolution remotely sensed images, the dimension of the ground object is a little bigand the details are blurred. While in the high resolution images, the dimension of theground object is small, the details are refined and it has clear relations with thesurroundings, which will lay the good foundation for the processing and analysis ofremotely sensed images. Therefore, the technology of high spatial resolution remotesensing has been applied in many fields, especially in the case of informationextraction which has showed its important application value.However, the complex and varied information, changeful object structure andcomplicated interference information in the high resolution remotely sensed imagemake the use of high resolution remote sensing image to extract the feature faceenormous challenges. When the methods based on pixels are used to process highresolution remotely sensed image, only the spectral information of the pixels isconsidered, the texture information, context, and the shape of the target object in theimage are ignored. This leads to the serious salt and pepper effect. The results ofinformation extraction are difficult to meet the needs of application for the highresolution remotely sensed image. Therefore, the characteristics of the highresolution remotely sensed image should be understood in-depth, and the suitableinformation extraction technology should be explored, which can improve its application value. Based on this, the principle of multiresolution segmentation wereexpounded, the scale parameter, spectral heterogeneity, shape heterogeneity andother factors were analyzed in this paper. The methods of edge detection werestudied, and the edge informations of the ground objects were integrated toparticipate in the multiresolution segmentation to create the image object layersassociated with practical objects. Based on the analysis of the characteristics of theobjects, and the main features of the study area were extracted by combiningthreshold classification with fuzzy classification. Lastly, moderate amount ofsamples were chosen to calculate the confusion matrix to evaluate the accuracy of theresults. Through the study, this paper obtained the following results:1) The results show that canny operator can obtain higher quality edge offeature on the edge detection by comparing the results obtained by differentalgorithms. Aiming at the existing problem of automatically extracting edge, somespecial difficult edges were extracted by artificial method. This can provide edge datafor multiresolution segmentation.2) The image object layers were created by image segmentation based on furtherstudy the principle of multiresolution segmentation technology and the importantimpact factors. The results show that the method can divided interferenceinformation and the adjacent pixels into the same homogeneous area, which caneffectively reduce or eliminate noise interference, and solve the problem of localheterogeneity in the image. The problem that "the same object has different spectra","different objects have same spectra" can get better solved, and the "salt and peppereffect" widespread in the results by the traditional methods can be overcame.Therefore, in the application of information extraction based on the high resolutionremotely sensed image, the multiresolution segmentation is undoubtedly a reliablenew method.3) The edge information was integrated in the object image segmentationaiming at the defects in the simple use of multiresolution segmentation. The optimumscale of the ground object was determined by analyzing the relation between thebiggest area of object and the scale. The results show that the quality of image segmentation has been greatly improved. It is quite clear that the edge is obviousbetween objects, the object is more effective and separable, and thus the accuracy ofinformation extraction has been improved.4) Through the systematic analysis of the image object features and practicalground object features, extracting the main features of water, residents, vegetation,road and other object. Using image object as the basic unit, using threshold and fuzzyclassification method to extract the ground object feature information, Finishing theextraction results, making thematic map of critical ground features in the study area.Finally，select the evaluation results after computing confusion matrix of propersample, which showes that all types of feature extraction with a high accuracy, theoverall precision of94.6%.The results showed that according to ground objectcombined with the use of two methods can get better extraction results. Based onground object information extraction method with multi-scale segmentation hasobvious advantages and good prospects in the high resolution remote sensing dataapplications.