基于序列图像的视觉定位

提出了一种静止摄像机条件下基于计算机视觉的工业机器人的定位算法。采用单目视觉系统,由线性摄像机模型推导出定位测量模型。通过序列图像,以一种改进的自适应混合高斯模型更新背景,运用背景差分法检测出运动目标,用CAMSHIFT算法实现对运动目标的跟踪,确定其位置和姿态,估计出运动参数,反馈给机器人指导其准确运动。实验结果表明,该方法快速、有效,能够满足视觉定位的实时性要求。

Visual Localization Based on Sequential Images

An industrial robot localization method based on computer vision is proposed under the condition of a static camera. A monocular vision system is introduced to deduce the localization measurement model by means of a linear camera model. Through the sequence of images, a moving object is tracked with CAMSHIFT algorithm, following detecting according to background difference method in which a modified self-adaptive mixture Gaussian model is adopted as the background updating method, thus localizing them and estimating their motion parameters in order to guide the robot. Experimental results show that the presented method is fast and effective, so it is able to meet the requirement of real-time visual location.