基于时间一致性局部颜色特征的无纹理3D物体实时跟踪

针对无纹理3D物体跟踪算法在复杂背景和运动模糊的情况下容易跟踪失败、跟踪速度难以达到强实时等问题,提出一种基于时间一致性局部颜色特征的3D物体实时跟踪算法.首先在物体3D模型投影轮廓法向搜索线上计算像素颜色的加权均值作为局部颜色特征,增强颜色特征在复杂环境中的表征能力,并对局部颜色特征进行时间一致性更新,剔除前景背景颜色相似的局部颜色特征,以避免相似前景背景颜色导致的跟踪失败;然后定义基于局部颜色特征的能量函数,并推导该能量函数的解析导函数;最后改进了优化物体姿态的高斯牛顿法,通过添加阻尼参数防止姿态优化陷入局部极值,提高姿态估计精度和跟踪速度.实验使用7组测试视频验证文中算法,结果表明,该算法能更有效地克服复杂背景和运动模糊的干扰,在未使用并行计算的前提下可实现强实时跟踪.

Real-Time Textureless 3D Object Tracking Based on Temporally Consistent Local Color Feature

Textureless 3D object tracking often fails because of cluttered backgrounds and motion blur,and slow tracking speed is difficult to achieve real-time tracking.To overcome these problems,this paper proposes a real-time textureless 3D object tracking algorithm based on temporally consistent local color feature.Firstly,color pixels are collected on the normal search lines that pass through the projected contour points of 3D model,and weighted mean of these pixels color is computed as local color features,which can enhance representation of color feature in cluttered backgrounds.The temporally consistent updating strategy is applied to the local color features,and local color features with similar foreground-background color are eliminated in the updating process to prevent tracking failure caused by similar foreground-background color.Then a novel energy function is defined based on the local color feature,and the differentials of the energy function with respect to the pose parameters are derived.Finally,the optimal pose parameters are searched using improved Gauss-Newton algorithm,in which a damping parameter is used to prevent convergence to a local minima,so that the accuracy of pose parameters and the tracking speed are improved.7 real video sequences are tested to verify the performance of the proposed algorithm.Experiments demonstrate that the proposed algorithm can overcome the influence induce by cluttered backgrounds and motion blur,and implements real-time tracking without relying on parallel computing techniques.