Research On Road Extraction From Midium And High-resolution Imagery

Extracting the road network in remote sensing images was an important part of remotesensing as well as survey and mapping, the extraction of road property was a basic content ofterrain information acquisition in relief mapping with essence scale, the research was regardedby remote sensing and survey and mapping all through the ages. However, in midium and highresolution images, road cannot simply be treated as linear objects, and subject to the impact ofvehicles, symbol lines of traveling crane on road surface as well as virescence strip and row trees,it was a hard work to extract road property information. On the analysis of the characters of theroad in many kinds of images, we mainly used region growing method, object-oriented mind andthreshold intersected ways to work on road information extraction, and the results and algorithmsunder different parameters were analyzed. In the end, with the knowledge of mathematicalmorphology we post-processed the results and extracted the center line of the road. The papermainly to complete the work:1. Filter the image by a template derived from Laplacion., and gained road results with alittle noise, based on this result using region growing and area statistic algorithm to extract road,wiping off the noise.2. Tried the road extraction with the application of object-oriented approach. On the basis ofmulti-resolution segmentation of road, we analyzed the characteristics of the road on the imageto sum up road features knowledge, worked on how to use these characteristics to achieveextraction of the road.3. By analyzing the genetic algorithm, discussed relationship between elements thatinfluence the road extracting precision and reliability, and find the rule to improve the fitfunction of the genetic algorithm. The road extraction results showed that this paper had workedwell.4. On high high-resolution images, we regarded road as areas with a certain width, and usedthe supervised classification means, and lead by road stylebook, receive good road area viaanalyzing the distance and threshold.5. Using the idea of mathematical morphology, and aimed at extraction specialty of mediumand high high-resolution images, we processed the results precisely, and gained the sideline andcenterline of road. The experimentation indicated that our technique ways had severalcharacteristics: high process speed, better unti-yawp ability and credible precision.