Three Efficient Object-Oriented Remote Sensing Change Detection Methods For Potential Safety Hazard Along High Speed Railway

With the rapid devolopment of high speed railway, the operating security attracts concerns of the whole nation. So the State Council required related departments to carry out safety check along high speed railways. One of the major contents is the environmental management situation along high speed railways. To detect the potential safety hazard in the environment quickly and offer technical supports to related departments to carry out the security check actions along high speed railways, this paper launched researches of fast object-oriented remote sensing change detection methods for potential safety hazard along high speed railways.This paper firstly confirms research contents according to research target: object-oriented change detection methods based on GIS data. Secondly, remote sensing image characters and typical automatic threshold segmentation technologies which involved in object-oriented change detection methods based on GIS data is systematacially surveyed. Thirdly, the efficiency of three existing object-oriented change detection methods based on GIS data are assessed. Fourthly, three efficiency object-oriented change detection methods are proposed. Then compare the efficiency and accuracy of the efficiency and existing ones respectively through two experiments. In the first experiment(image size:1154X5064), the time-consuming of six methods above turns(corresponding to the order in text) for98s,95s,102s,93s,86s,75s. Overall accuracy turns for80.65%,82.70%,77.57%,87.21%,88.69%,81.69%. Misdetection rate turns for8.64%,7.43%,10.53%,1.86%,2.43%,7.43%. False positive rate turns for10.71%,9.87%,11.90%,10.93%,8.88%,10.88%. In the second experiment(image size:16724X3564), time-consuming turns for152s,143s,170s,140s,129s,118s. Overall accuracy turns for74.19%,78.47%,71.96%,80.56%,83.41%,77.87%. Misdetection rate turns for11.43%,9.83%,12.75%,3.75%,4.87%,10.12%. False positive rate turns for14.38%,11.70%,15.29%,15.69%,11.72%,12.01%. Through these results we can see that Differential gray gradient-Maximum entropy method is the most efficient and Modified average gray vector-Otsu method takes second place. In accuracy aspect, the overall accuracy and misdetection rate of Average gray vector-Otsu method and Modified average gray vector-Otsu method are superior to the existing ones, in which Modified average gray vector-Otsu method has the best overall accuracy while Average gray vector-Otsu method has the best misdetection rate. These results confirmed that the new methods are efficient and practical. Finally, this paper applies three efficiency change detection methods to the change detection along Zhengzhou-Xi’an high speed railway(image size:78161X7103), and the time-consuming turns for2h44min35s,1h35min56s,1h28min45s. Overall accuracy turns for80.97%,82.72%,77.57%. Misdetection rate turns for3.16%,4.36%.10.75%.False positive rate turns for15.87%,12.92%,11.68%. The efficiency is respectively3.31%,43.64%,47.86%higher than character correlation coefficient-maximization objective function method which is the most efficiency one in three existing methods.The results above show that the faster object-oriented change detection methods presented in this paper have a certain use value in the application of change detection along high speed railway and updating high speed railway database, and provide reference to further follow-up researches.