Research On Forest Land Extraction Methods In Remote Sensing Images

Woodland is an important part of forest and an important indicator of ecological environment,so it is one of the primary and important tasks to grasp woodland resources and monitor woodland changes in real time.In recent years,with the development of remote sensing technology,remote sensing images are often used to monitor the real-time dynamic changes of forest land.Remote sensing technology can obtain woodland information quickly and periodically,providing a convenient means for woodland dynamic statistics,which is also the development trend of monitoring the dynamic changes of woodland areas.Woodland information statistics is generally the process of spatial identification and spatial positioning of woodland areas and quantitative calculation of woodland areas,in which accurate segmentation of woodland areas in remote sensing images is an important basis for woodland area statistics,but the types of features in remote sensing images are complex,and mis-segmentation and under-segmentation are easy to occur when extracting woodland information,resulting in lower segmentation accuracy.The fuzzy clustering algorithm is fuzzy and can better deal with the uncertainty problem,which is widely used in images with complex features like remote sensing images.Therefore,this paper uses multiple features of the image and the improved fuzzy clustering algorithm to extract the forest land information in remote sensing images.The main work of the thesis is as follows:(1)Aiming at the problem that the traditional clustering segmentation algorithm only considers a single grayscale feature of the image,which leads to less than ideal results when the algorithm is used for remote sensing woodland extraction and is prone to misclassification and omission,an En FCM remote sensing image woodland extraction method based on PCA multi-feature fusion is proposed.Firstly,texture features and edge features of the image are introduced on the basis of grayscale features of the image;secondly,the PCA principal component technique is used to assign different primary and secondary importance to each extracted feature and generate a multi-feature fusion image,and the fused features are used to measure the differences between pixels instead of single grayscale features;finally,the improved En FCM method is used to calculate the similarity between pixels and clustering centers to achieve woodland extraction.The experimental results show that the method can effectively improve the accuracy of forest land segmentation in remote sensing images compared with other clustering algorithms.(2)Aiming at the problem that the traditional clustering algorithm only considers the local information of the image,and the remote sensing image may be disturbed by noise in the process of acquisition and transmission,which causes the pixel value and spatial structure of the image to change,making the local similarity measure have errors to the extent of over-segmentation or under-segmentation,a method of extracting forest land from remote sensing image of FGFCM that integrates local information and non-local information is proposed.Firstly,the non-local information of the image is introduced on the basis of the local grayscale information and spatial information of the image;then the non-local information influence factor is determined adaptively by the strength of the non-local noise on this basis,and a new linear weighted sum image is obtained by combining the local and non-local information of the image with this influence factor;finally,the clustering operation is implemented on the grayscale histogram of this filtered image.The experimental results show that the improved method can better preserve the image edge details while improving the image segmentation accuracy.