自然环境中显著性多路标提取算法的研究和改进

Algorithm and Improvement for Extracting Multiple Salient Landmarks from Natural Environment

自主移动机器人利用视觉传感器在自然环境中进行探索,因为没有任何先验信息,只有提取自然环境中比较显著的物体作为路标。因此,自然路标提取是对未知环境开始探索的第一步,也是至关重要的一步,因为它直接影响到移动机器人后续的定位导航的完成,文中融合亮度、纹理1、颜色等信息,结合中心上-中心外机制得到显著性区域,并采用分块聚类方法跟踪完整提取多显著性区域,实验结果表明本算法对显著性区域具有较好的检出能力,并且基本完整提取多个显著性自然区域路标,能够适应远近尺度、旋转、视角变化及一定的光线变化等自然识别的要求。

When the mobile robot explores in the natural environments with visual sensor,it has to detect the nature objects that are relatively salient in nature environments as landmark.So nature landmark detection is not only the primacy but also most important in unknown environment exploration,because it directly influences the completion of navigation.A visual saliency region detection system blending intensity,color,texture and centre-on-off mechanism is presented in this paper.With the method,the robot can detect several integrity saliency regions by way of clustering method.Some experiments are executed to verify the algorithm in detecting saliency region,picking up several saliency nature landmarks and repeatability including scale,rotation and viewpoint invariance.