Research And Application On Key Technologies Of Multi-source Remote Sensing Image Railway Environmental Monitoring

The rapid development of China’s railway transportation industry has greatly improved the level of people’s travel.However,the railway will change the original geographical landform,reduce the diversity of plants and animals,and lead to a series of ecological damage problems such as the reduction of environmental protection vegetation coverage rate and the acceleration of soil erosion.The traditional ground monitoring method is single and has a long period,which makes it difficult to truly reflect the ecological changes around the railway construction.Satellite remote sensing has the characteristics of wide detection range,short reentry period,low cost and few restrictions,which can effectively make up for the deficiency of monitoring.With the gradual improvement of the satellite system,a large number of multi-sensor,multitemporal and multi-spectral remote sensing data,namely,multi-source remote sensing images,can be obtained.Multi-source remote sensing data not only retains the advantages of single source data such as wide amplitude,easy data preservation and short time consumption,but also forms the complementarity and cooperation among different data.Therefore,this thesis uses multi-source remote sensing images to monitor changes in environmental protection conditions around the railway,assist supervisory personnel in timely monitoring,and improve the accuracy of supervisory data.The main research contents are as follows:(1)Research on multi-spectral and panchromatic image fusion based on sparse representation and salient map.By enhancing the image fusion effect,the subsequent change detection algorithm can achieve better results.In view of the problem that background information and texture information are easy to be lost in image fusion,this paper adopts online dictionary learning based on sparse representation to perform low-frequency subband coefficient fusion,so as to retain spectral and background information of the two kinds of data,reduce data redundancy and improve algorithm efficiency.Gaussian mixture model is used for clustering to generate salient map,and normalized high-frequency coefficient fusion is carried out through spectral contrast of salient map to enhance the detailed information.Finally,the experimental results show that the improved fusion method can retain the geometric texture information of panchromatic image as much as possible while maintaining the spectral information,so as to provide more relevant data information for the subsequent practical application.(2)Research on remote sensing image automatic monitoring technology of railway environmental protection based on interval type-2 fuzzy C-means clustering based on improved firefly.Remote sensing image change detection can dynamically monitor the environmental changes along the railway.However,the complexity and fuzziness of remote sensing image will interfere with the results of image change detection.The interval type-2 fuzzy C-means clustering algorithm is introduced to solve this problem.But the randomness of the algorithm parameters makes the detection results unstable.In addition,the imaging uncertainty of remote sensing image will affect the accuracy of change detection.Based on this,an improved firefly algorithm is used to complete the parameter determination method of interval type-2 fuzzy Cmeans clustering algorithm to realize remote sensing image change detection.The local optimal solution is used to select more refined candidate solutions of firefly algorithm and the variable step size factor is introduced,to search more accurate fuzzy factors.Combined with the fuzzy factors obtained from the optimization,the interval type-2 fuzzy C-means clustering is performed.Finally,the clustering centers are optimized by the weighted Karnik-Mendel algorithm based on the compound trapezoid rule,and the change types are determined according to the maximum membership principle.The experimental results show that the improved method improves the accuracy and stability of the change detection,and can further monitor the data and evaluate the relevant indexes of the change detection results of environmental protection features along the railway.(3)Design and implementation of multi-source remote sensing railway environmental monitoring system.The improved algorithm is encapsulated and integrated,and the system is designed and realized,including image preprocessing,image fusion,environmental protection object change monitoring and evaluation,environmental protection early warning and other modules.By adding interfaces,the function of the system is expanded,and a railway environmental protection monitoring system with convenient operation and complete functions is finally realized.