提出了基于轮廓模型的复杂背景弱纹理目标单应优化方法。算法在随机抽样一致(RANSAC)框架下实现了初始变换的求解，通过优化法向距离实现了单应的优化求解。为了快速稳健地求解初始单应，算法随机选取三条满足一定几何约束的直线段进行假设变换关系的求解，通过选取使得投影误差最小的变换关系作为单应初值。为了解决复杂背景条件下模型-图像对应错误引起的优化失败问题，在模型-图像点匹配阶段，算法为每个采样点保留多个图像点对应，同时在对样本点进行加权过程中，该算法综合考虑了样本点自身的属性和样本点同周围点的关系，有效提高了稳健性。实验结果表明：该方法能够实现复杂场景目标单应的优化求解，相比传统的方法，该方法能够有效克服复杂背景的干扰，实现弱纹理目标单应的稳健估计。

A method based on contour model tracking to estimate the homograhpy for the textureless object in the clutter scene is proposed. The initial estimation of the transformation is obtained in the framework of random sample consensus (RANSAC). The optimized homography solution is obtained by minimizing the normal distance. To calculate the initial transformation quickly and robustly, random three line segments conforming to the certain geometry constraint are utilized to solve the assumptive transformation relation. The transformation relation with the minimal errors are picked out as the homographic initial value. To overcome the issue that the mismatches of the model and image lines in the complicated background may lead to the failure of the homography optimization, in the process of matches of the model and image points, multiple image points matches are retained for each model sample point. In the weighting process for sample point, due to the use of the property of the sample point as well as the relation to the neighbor points, the robustness of method is enhanced effectively. Experimental results show that the proposed method can realize the optimized solution of homography in the complicated background. Contrast with the traditional method, the proposed method can overcome the influence induced by the complex background effectively and optimize the homography parameters for textureless objects successfully.