Vector-data Assisted Methods For Detection Of Change Along High-speed Railways With High-Resolution Remote Sensing Images

Applying the remote sensing technology for dynamic monitoring of land resources has become an effective approach to know about and conduct macro control on the region’s economic activity. With the rapid development of the computer technology and remote sensing technology, especially the kind with high spatial resolution, it has become a possibility that using this technology to monitor the land utilization, detect the hidden danger along the high speed rail line and provide a scientific basis and technical support for the railway sector to carry out high-speed rail safety inspection. Therefore, with the application of high-resolution remote sensing image, this paper studies change detection method on the land feature and environmental characteristics along the high-speed railway, and conducts further discussion and analysis.Based on the summary of related studying achievements about the remote sensing image change detection and according to the land feature and environmental characteristics along the high-speed railway, this paper mainly studies the change detection method in combination with GIS vector data. Considering the present research status, there are many change detection methods using GIS vector data, but their effects vary from the needs of different change detection and the characteristics of experimental data and there are no comparatively mature change detection framework and system on the remote sensing in combination with GIS vector data. Therefore, on the basis of summarizing the general idea and process of these methods, this paper discusses several existing change detection methods based on GIS vector data and taking image segments as research object. Meanwhile, image segment spatial histogram feature is introduced and a new change detection method using GIS data is proposed. As for these methods, detection experiments are conducted on the image scenes with different land features and discussions about the practicability of each method on different scene image are made as well.Through the experiment, found that at the second phase images, part image segments obtained in using vector data split may lead local changes, but the detection process of image segments will be judged as all change or not change. Based on the above, this paper proposes a change detection method judging the preliminary change (the whole variation of image segments, image segments all not change, local variation of image segments) of the image segments in second period through analyzing the grey degree distributing of image segments, and judging the image segments with local variations by building and obtaining the local variation coefficient of image segments. Finally, this method is proved effective through experiment and the the experimental results suggest that the method, with high practical value, can detect some local variation areas, which cannot be done with GIS data direct segmentation detection method. On test precision, in the first set of experimental data, two methods of using detection image segments compared with direct use vector data segments result tested increased by 1.18% and 2.69% respectively; In the second set of experimental data, the overall accuracy increased by 2.93% and 0.54% respectively; In the third set of experimental data, the overall accuracy increased by 3.36% and 3.77% respectively; In the fourth set of experimental data, the overall accuracy increased by 2.11% and 1.20% respectively.