Object Extraction Of Waterway And Port Using Line-patch-fused Image Analysis Technique

As a rapid and effective means of earth observation,remote sensing technology has been widely applied to the daily life,however,high resolution remote sensing image processing technology is not mature enough. In recent years, the object-oriented image analysis technique is used in RS image,and it work better than the traditional image analysis method based on pixel.Now the object-oriented image analysis technique only based on image segmentation.It not only cannot use the features(spectrum, shape, texture, spatial relations and so on)for image analysis comprehensively and effectively,but also cannot process the over-segmentation phenomenon.Aiming at the existing problems,we propose a kind of object oriented image analysis method which from the perspective of line-patch fusion.Combined with characteristics of the typical artificial structures in port and channel,realized the extraction of Bridges and dam from high- resolution RS image.Our work includes two parts as follows:(1)we use the segmentation of RS imagery based on hard-boundary constraint and two-stage merging to get segments,and we use a straight line extraction method based on features fusion.Combining the feature of ground objects,establish space relationship of lines and patchs.This space relationship can solve over segmentation well.By using this space relationship, patchs of same object can be related.(2)We propose a new extraction method of bridgand dam which based on line-patch fusion.Frist use RS-image classification to get the scope of water; then combining with the space relationship of target,water and ground objects, the lines of target is extracted by using the water-patchs;last, the target-patchs is extracted by using the Characteristics and space relationship of the targets. Experimental results show that the extraction method of this pepar can extract bridge and dam effectively from high spatial resolution remote sensing images.