Research On Image Processing Method In Geographical National Conditions Using The Remote Sensing

The first Geographic National Conditions Investigation has been carried out from 2013 to 2015. It is a Groundbreaking work. Its heavy task, high technology, strong professionalism,wide work scope and many departments involved are unprecedented. High-resolution satellite images used in the Geographic National Conditions Investigation are multitemporal and multisource. How to select and process the images determines the Quality, precision and progress of later work. Therefore, study on remote sensing image processing method for the Geographic National Conditions Investigation has an important significance. The results of this paper are as follows:(1) According to the main types and characteristics of high-resolution remote sensing image, this paper proposed the technical methods and processes of making orthophotos with high resolution remote sensing image, including data preprocessing, ortho-rectification,registration of multispectral image and panchromatic image, fusion, color adjustment, mosaic and cutting etc. Among the rest, this paper focused on the impact on the classification of image fusion. It was found that image fusion had a large effect on classification extraction.(2) This paper summarized and analyzed the basic characteristics of roads, waters,vegetation and buildings during the automatic classification of high resolution satellite images.Based on object-oriented method, an extraction experiment of roads, waters, vegetation,buildings had been done. This experiment studied on the extraction method based on information index. Compared to the extraction result which only used NDWI, use improved rulesets to extract water can receive a good result. The relative error, average relative error,weighted relative error of total area of waters were relatively small. The average area ratio increased from 71.15% to 96.52%, which achieved a better classification results and effectively improved the classification accuracy. In connection with the images only had RGBthree bands, this paper presented four rulesets of vegetation extraction and did classification accuracy assessment. The result showed that the average area ratio of the VI’ classification could reach to 88.41%.(3) According to the visual interpretation types of land cover, this paper discussed the technology based on eCognition interpretation of human-computer interaction. Segmentation scale assessment tools ESP(Estimation of scale Parameter) was used to obtain the optimal segmentation scale of different objects and levels in high-resolution images. The effects of visual segmentation and small spot vector matching determined the optimal segmentation parameters, which effectively avoided the phenomenon of over-segmentation and under-segmentation. Thus, it could provide the basis for the parameters of image segmentation while effectively Reduce the time of image segmentation.(4) This paper also produced the vector data smoothing model processor based on ARCGIS software, which effectively solved the output vector data presenting jaggies after classification while achieved the purpose and precision of rapid collection of land cover data.The innovation of this paper is to propose a set of technology roadmap for human-computer interactive vectorizing of the Geographic National Conditions Investigation data. It greatly reduce manual workload and improved daily productivity. It also provided practical technical basis for the extraction of high-resolution images’ information and had reference value for sustainable monitoring of the Geographic National Conditions in the future.