论文部分内容阅读
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 object detection and recognition is a very active area in machine vision at present. Based on the geometrical and optical properties of the medium-pitch bridge, a knowledge base of the medium-pitch bridge is formed, and an effective mid-span bridge identification algorithm is proposed. The algorithm solves the problem of low contrast and multi-interference image enhancement. According to the statistical recursive Otsu segmentation, the false segmentation of water area is effectively solved. The knowledge of middle distance bridges is used to judge the real water areas on both sides of the bridge. The method to set the confidence of mid-span bridges based on knowledge is proposed. Experimental results show the effectiveness of the proposed method.