【作者】 杨渊； 李寒松；

【Author】 YANG Yuan, LI Han-song (35th Institute of 3rd Academy, China Aerospace Science and Industry Corporation, Beijing 100013, China)

【机构】 中国航天科工集团第三研究院第三十五研究所；

【摘要】 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。更多还原

【Abstract】 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.更多还原