Research Of The Technology Of Road Damage Information Extraction Based On SAR Imagery

As an important transport facilities, road acts a great role in daily life, economic, military and other fields, thus it is necessary to do some efforts in road damage information extraction, particularly, the research may provide some scientific evidences in disaster investigation, disaster relief, which is significant. The methodology of road damage information extraction based on SAR images aims at utilizing the characteristics of all-time, all-weather of SAR image to achieve timely and effective road damage recognition, compensating for the lack of optical image, it is divided into two parts, road extraction and the extraction of road damage.Firstly, This paper analyzed the image characteristics and road feature extraction methods in this part, and then according to the characteristics of mountain scenes and image characteristics, the selection of target features and road extraction method is done. For more details, firstly an appropriate image enhancement algorithm is selected to filtering the basis image, after that an edge detector is employed to extract the edge strength information of road, and based on the gray information, the edge intensity information and the edge direction information, the author utilized the particle filter algorithm for road object recognition to get an initial contour point sequences according to the idea of track before detection, which overcomes robustness shortage of the globally auto connection method when employed in mountain environment directly. Furthermore, in order to make the finally extracted road lines and real roads have better compliance, this paper take advantage of Snake method to make initial contour point sequence acquired from particle filter result more precisely corrected, which outputs the road extraction result. Experimental results when applied in Radarsat2 images show that the above method has good accuracy, which can meet the requirements of the following road damaged information extraction procedures.Secondly, road damage information extraction section, common types of road damage and road damage characteristics variation in SAR images were analyzed, from which the changes of edge feature and gray feature are selected as the proof of damage detection, and a new road damage information extraction method in SAR image is proposed. The implement ways as followed, the co-registration operation is done in dual-temporal SAR images, and then built the road edge damage index model and gray damage index model in the slave image to be detected, based on the road lines extracted in the basis image. and then analysis the statistical characteristics of the two indices with the samples from the road areas in basis image, from which an automatic extraction method to determine the damage thresholds of the indices is proposed according to the method CFAR principle, discriminate the edge damage index and the gray damage index with these thresholds, combining the discrimination results to obtain the extraction results of road damage.Finally, this road damage information extraction method was verified in dual-temporal Radarsat2 images, the availability and effectiveness of this method is demonstrated with the comparison between the results of detected image and the analysis of optical data, the artificial visual interpretations and the results obtained from the basis images with the same method.