Research And Realization Of Bridge Recognition

In recent years, the man-made target identification and detection for natural scenery cause the majority of scholars more and more attention, and it has very important value in military and civilian areas. Bridges are used as one of the typical man-made objects, identification, and it has become a hot research topic, but from the references at home and abroad, it is still does not have an identification methods which has good effect and be used a wide range now.In this paper, through research of target recognition algorithms of bridges over water in visible remote sensing image, for the existence of some of the shortcomings and deficiency of the algorithm, a segment method based on multi-scale and multi-structuring elements was presented.In the usual bridge recognition algorithms, building knowledge model of bridges and its scene at first, through gray value of the linear tension having a choice for the image, to enhance the contrast of interesting in the gray value range, and then using the method of threshold segmentation to extract water area, taking full advantage of spatial relations between rivers and bridges, according to the characteristics of the river across the image borders, taking closed operations connected with the river, through the detection rivers region of the connecting, so enabling the detection and recognition for bridges. This paper used multi-scale multi-structural element segmentation algorithm based on the premise of the bridge knowledge, adopting a large-scale segmentation algorithm in remote sensing image segmentation algorithm at first, extracting the image part of interest, then conducting small-scale image segmentation algorithm to image of separating out, extracting the outline of the target, which is an innovative point of this paper.This paper proposed a new line segment detection algorithm based on Beamlet transform in bridge detection and identification, it achieved target recognition of bridges over water in remote sensing image through selecting the appropriate model. Experimental results show the proposed algorithm can quickly detect whether contain a bridge in remote sensing images, thus avoiding the blind operation, while providing accurate positioning of the bridge. In addition, the algorithm has some advantages, such as broader scope of application, lower false alarm, lower missed error, more accurate, and higher speed.