Change Detection Based Land Cover Type Map Updating Method Using Remote Sensing

In recent years,with the rapid development of remote sensing satellite in China,remote sensing image has been widely used in land cover measurement,monitoring and mapping.Forest is an important part of land cover type.For a long time,the mutual transformation relationship between forest and other land cover types is an important research content for obtaining forest cover change information and land cover types information in forest resource monitoring.In recent years,the production methods of land cover data related products have been gradually enriched,and the products have greatly improved in spatial resolution and classification accuracy.However,due to the complexity and dynamics of land cover types affected by human activities and natural factors,the production of land cover products still faces many challenges.On one hand,the acquisition of land cover type information requires stable remote sensing data sources to provide sufficient data;On the other hand,it is also necessary to minimize the cycle and cost of data update under the premise of ensuring the accuracy of data products.Compared with other methods,the method based on change information can reduce the workload and at the same time keep the consistency of updated data and historical data to the largest extent,and thus become an important method for updating land cover type maps.Based on the above background,the following work was carried out to explore the update method of land cover type map based on change detection:(1)This paper optimizes the method for calculating the characteristics of change detection and uses a multivariate alteration detection(MAD)method based on canonical correlation analysis.Through analyzing the characteristics of MAD components corresponding to different types of land cover changes,it is determined that The distribution of land cover change information is disordered,and this feature can be used to eliminate "pseudo-changes" due to different crop types or growing seasons in the "cultivated land" type.By this method,a more accurate MADsyn difference image is constructed.(2)This paper uses Markov Random Field(MRF)for binary segmentation of MADsyn difference images to achieve higher accuracy of land cover change.By comparing Otsu method with the MRF binary classification method,the effect of neighborhood information in the process of determining the change threshold were studied.The results show that the accuracy rate of the change detection results obtained by Otsu and MRF model methods are respectively 70.21% and 76.69%.The false alarm rates are 30.12% and 23.57% respectively.The missed detection rates are 8% and 6% respectively.The study verified the effectiveness of the MRF binary segmentation method used in this study and obtained a high accuracy of regional distribution detection.(3)A method was developed to establish a classification rule set using classification features to effectively identify the land cover type of the change area in the later period.The experimental results show that the combination of NDVI and texture difference features can well distinguish the change directions such as increased vegetation cover,reduced vegetation cover,increased water body,and increased construction land.The accuracy of land cover coverage obtained by the study reached 81.25%.(4)The statistical analysis results of the results of the spatial variation of land cover types in the study area indicate that the net growth area of forest land in the study area was 426.94 km 2 in 2013-2017,and the growth rate was 2.54%.The major change types were grassland-to-forest land conversion,and the conversion The area is 450.5km2,and other areas also have small-scale mutual transformation.The main drivers of change are the implementation of forestry policies and forestry projects.In summary,this paper develops a land cover type map updating method based on change detection,which has a relatively high efficiency and accuracy of land cover type map updating.It is conducive to the application department to timely and comprehensively grasp the dynamic changes of forest resources in large areas.