Research On Extraction Method Of Forest Land Resources Information From Remote Sensing Images

Forest land resources are valuable natural resources for human beings,and counting the area of forest land areas are one of the main steps to understand the basic situation of forest land resources,and extracting forest land areas from remote sensing images are of great significance in land and urban planning and management.Traditional forest land extraction methods usually use manual field surveys,which are not only time-consuming and labor-intensive,but also take a long time to measure,are slow speed and have a low accuracy rate.The remote sensing technology can obtain remote sensing data quickly,accurately and in real time,which is very helpful for the investigation and research of forestry resources.The use of remote sensing technology for classification and ecosystem target identification in complex forest areas can provide more information for natural resource management,land planning and other studies.Therefore,in order to overcome the difficulties of traditional measurement methods,this thesis analyzes and identifies forest land by using remote sensing technology.The types of ground objects contained in remote sensing images are complex,and errors and omissions are prone to occur when extracting information from forest land,resulting in low accuracy.Therefore,this thesis uses morphological processing and an object-oriented method based on texture features to extract forest land from remote sensing images.The main work of this paper is as follows:(1)This thesis summarizes the research status and development significance of forest land extraction methods,adaptive morphology and texture features in remote sensing images at home and abroad,and introduces the basic theory of remote sensing image segmentation and classification.(2)For improving the segmentation accuracy of forest area in remote sensing images and solving the problem of inaccurate segmentation of forest area in remote sensing images caused by factors such as coverage and noise,a remote sensing image forest area extraction method based on GAN(General Adaptive Neighborhood,GAN)adaptive morphological composite filtering was proposed.First,GAN morphological erosion and expansion operations were constructed with GAN adaptive structure elements,through which GAN morphological open and closed operations were derived.Secondly,the GAN morphological composite filter was constructed to fill the holes with insufficient coverage in the forest land area and reduce noise interference on the image.Finally,the gray-scale window slicing method was applied to segment the remotely sensed forest land images.Through experimental simulation,the method could effectively improve the forest area’s segmentation accuracy and make a complete segmentation of the forest area of remote sensing images.Compared with other remote sensing image segmentation methods,the segmentation accuracy of this method is higher than that of other methods.(3)The changes in forest resources are time-sensitive,and to understand the changes of forest information timely and accurately,as well as for the traditional pixel classi fication method cannot take into account the spatial information of remote sensing images,which leads to the problem of misclassification and loss of ground objects.This thesis performs forest land recognition on remote sensing images using texture features with object-oriented classification.By selecting remote sensing images of the Xihu District of Hangzhou as the study area,the images are classified statistically by adding texture features.Firstly,the Gray-Level Co-occurrence Matrix(GLCM)texture extraction is performed on the study area:the optimal window size and the optimal texture feature index are determined;then,the extracted texture features with the object-oriented K-Nearest Neighbor(KNN)classification method to classify and identify the forest resources of remote sensing images,and compare the results with the traditional pixel-based classification method and the classification method without texture features;finally,the transposition matrix is generated to analyze the changes of forest land.The experimental results show that the object-oriented KNN classification with texture features is better than the object-oriented KNN classification without texture features,and the overall classification accuracy with texture features reaches 94%.The Kappa coefficient reaches 0.92,the area statistics of the classified images are performed,and the change of forest land is analyzed.This method can provide an effective way for forestry information extraction.