In this paper, for the remote sensing images acquired under the GIS model, the wavelet thresholding denoising method is used to reduce the noise to improve the quality of computer images, combined with the image filtering method to complete the grayscale preprocessing of the image, and fused different operators to extract the edge features. To improve image alignment and decrease the amount of splicing calculations, a proposed raster DEM data fusion and splicing algorithm utilizes image feature information. An experimental environment has been created for preprocessing computer vision images by comparing peak signal-to-noise values at different levels to find the best parameters for wavelet thresholding denoising. The feasibility of the preprocessing method in this paper is tested by placing unpreprocessed and preprocessed images in DeepLabv3 and DAM-DeepLab models for training purposes. Test the effectiveness of this paper’s algorithm for fuzzy image target feature description by simulated image edge extraction process. Real-time and effective evaluation of image stitching verifies the reliability of the algorithm. The data show that the splicing results of the four different methods under different scene images are 6.6932 for the global transform algorithm, 6.6831 for the APAP algorithm, 6.6449 for the AANAP algorithm, and 6.6948 for the algorithm of this paper. This paper’s algorithm has seen an improvement in average information entropy by 0.16%, 1.17%, and 4.99% when compared to other algorithms. The theory of graphics application technology within GIS is enriched through the study of computer graphics processing technology within the GIS model.
Full Text Finder