Research On Remote Sensing Image Reconstruction And Segmentation Algorithm Based On Compressed Sensing And Swarm Optimization

With the continuous development of remote sensing technology,remote sensing image processing has been widely applied in many fields,such as geological exploration,environmental monitoring,and agricultural production.Reconstruction and segmentation of remote sensing images are important research directions because they can extract useful information from images for subsequent analysis.However,traditional image processing methods cannot meet the demands due to the high-dimensional characteristics,large data volume,and information complexity of remote sensing images.To address these issues,this paper studies a comprehensive process for remote sensing image processing,which involves transmitting and reconstructing remote sensing images before segmenting them.The specific details are as follows:Firstly,for the high-dimensional characteristics and unknown sparsity of remote sensing images,traditional greedy algorithms require the sparsity of signals as a known condition and cannot adapt to the high-dimensional characteristics of remote sensing images,resulting in poor reconstruction accuracy.To solve these problems,this paper proposes an adaptive dualthreshold matching pursuit algorithm based on variable step size backtracking strategy.Firstly,a suboptimal atomic set is selected through two adaptive thresholds.Then,through the variable step size backtracking strategy,atoms are selected twice to obtain the optimal atomic set.The algorithm can improve the complete reconstruction rate of signals in the case of unknown sparsity,while the variable step size backtracking strategy can effectively reduce the complexity of the algorithm.Through reconstruction simulation experiments,the reconstruction accuracy of one-dimensional signals can reach 100% under certain conditions,and the reconstruction speed is relatively improved.For two-dimensional image signals,the image quality of reconstructed remote sensing images using this algorithm is relatively enhanced compared to other algorithms at different compression ratios,which demonstrates the accuracy and effectiveness of the adaptive dual-threshold matching pursuit algorithm based on variable step size backtracking strategy.Secondly,due to the large amount of data and complexity of remote sensing images,image segmentation of remote sensing images will produce problems such as indistinguishable foreground and background,missing segmentation information,inaccurate pixel classification,and the need for manual threshold selection,which will seriously affect the visual expression effect and required information of images.Based on the above problems,this paper proposes an image segmentation algorithm based on group optimization.Firstly,this paper optimizes the Harris Hawks Optimization(HHO)algorithm by introducing elite reverse learning to improve its shortcomings in easily falling into local extremes and slow convergence speed when dealing with some complex,multidimensional optimization problems.Then,the improved Harris Hawks Optimization is combined with the traditional minimum cross-entropy threshold segmentation algorithm to solve the problems of manual threshold selection and inaccurate threshold selection,as well as to enhance the effect of image segmentation.The test results of remote sensing images show the accuracy and effectiveness of this segmentation algorithm.