Research On Application Of High Resolution Remote Sensing Image In Forest Land Change Detection

The National Forest Change Survey is based on the national forest land "one map",the forest land area,the forest land protection and utilization status,etc.are investigated and analyzed,it is to enhance the forest land supervision ability,strengthen the forest land protection utilization management,deepen the national and local government macroscopic decision-making management important foundation and the support.The important foundation and support.Traditional woodland changes mainly rely on ground-based surveys,artificial mapping of changes in borders,and problems such as large workload,low efficiency,and long cycle time.Therefore,traditional methods have been unable to adapt to the actual situation of the current forest land change investigation.Texture features have good spatial structure characteristics and are important information sources in image analysis.How to effectively use the texture features of the image to assist in the detection of changes in forest land has become the focus of current research.In this paper,we use the GF-1 image data covering Bamiantong Forest Farm in Muling City of Heilongjiang Province as the research object,and analyze and discuss the two different texture extraction techniques through the gray level co-occurrence matrix and two-dimensional Gabor filter respectively.The experimental comparison analysis shows that the two methods that use the texture information extraction method as auxiliary change detection are reliable and feasible.The specific research content is as follows:(1)A texture information extraction method based on a gradation co-occurrence matrix is used.For the more texture measure information filtered out by the gray level co-occurrence matrix,it is analyzed and screened.Through the study,it is found that there is a large correlation between different measures.For this reason,the selection of measurement indicators is analyzed and discussed.(2)Using two-dimensional Gabor filter to extract texture information technology.The two-dimensional Gabor filter has a very good ability of analysis and retrieval in the time domain and frequency domain.The design effect of the filter is similar to the biological human eye recognition technology,so it has multi-scale and multi-angle analysis features.Through the parameter selection and discussion of the Gabor filter,a set of parameters suitable for the study area was obtained.(3)The texture information extracted by the galactic co-occurrence matrix and the two-dimensional Gabor filter is used to assist the change detection of the GF-1 image.Through experiments,it is found that the two kinds of texture information extraction methods assist in image change detection and can obviously improve the accuracy of change detection.The total accuracy of the GF-1 image added with the gray level co-occurrence matrix contrast texture information and the texture information added to the two-dimensional Gabor filter to assist woodland change detection reached 83.44%and 83.030%,respectively,higher than that without adding texture information for woodland change detection.The total accuracy is 78.48%.