地物背景下中距桥梁目标检测与识别

Detection and recognition of middle-range bridge under the earth background

基于知识的目标检测与识别是目前机器视觉里非常活跃的一个领域.依据中距桥梁的几何及光学特性形成中距桥梁知识库,并依此提出了一种有效的中距桥梁识别算法.该算法解决了低对比度、多干扰图像的增强;根据统计意义上的递归 Otsu 分割有效地解决了水域误分割;利用中距桥梁的知识来判断桥梁两侧的真实水域,并以该水域来定位桥梁;提出了基于知识的中距桥梁置信度设置方法.实验结果表明了该算法的有效性.

Knowledge-based target detection and recognition is a very active field in the present computer vision research.An algorithm is presented which could effectively recognize the middle-range bridge based on its knowledge formed through the abstraction of its geometry and optics features.This algorithm performs a method to enhance image with low contrast;uses the recursive statistical Otsu algorithm to effectively reduce the possibility of error segmentation of water regions;estimates the true water regions beside the bridge by searching the middle-bridge knowledge bank so as to realize the registration of middle-bridge target.Finally,it gives a thorough description of how to set confidence of middle-bridge target.Simulation indicates the effectiveness of this algorithm.