CAJ | 학술논문

摄像机矩阵估计是机器视觉的一个重要问题。在2范数误差代价函数模型下，最小二乘法简单而有效，但因误差代价函数非凸，容易陷入局部最优。在无穷范数误差代价函数模型下，凸优化方法理论上可以获得全局最优，但计算效率较低，其计算耗时随着问题规模的增大而急剧增加。现代优化论中的增强连续禁忌搜索（Enhanced continu-ous taboo search，ECTS）方法具有逃离局部最优的优良性质，因此本文在2范数误差代价函数模型下提出一种针对摄像机矩阵估计的ECTS算法。在ECTS置信区间序列构造及最大置信区间选择环节，本文提出了一种非迭代的方法获取包含全局最优解的凸包。在增强禁忌搜索环节，本文提出了一种基于伪凸函数的候选解邻域构造方法。同时，给出了本文算法以概率1收敛于全局最优的理论证明。对虚拟场景和真实场景的实验结果表明本文算法可以快速获取摄像机矩阵估计的全局最优解。
섭상궤구진고계시궤기시각적일개중요문제。재2범수오차대개함수모형하，최소이승법간단이유효，단인오차대개함수비철，용역함입국부최우。재무궁범수오차대개함수모형하，철우화방법이론상가이획득전국최우，단계산효솔교저，기계산모시수착문제규모적증대이급극증가。현대우화론중적증강련속금기수색（Enhanced continu-ous taboo search，ECTS）방법구유도리국부최우적우량성질，인차본문재2범수오차대개함수모형하제출일충침대섭상궤구진고계적ECTS산법。재ECTS치신구간서렬구조급최대치신구간선택배절，본문제출료일충비질대적방법획취포함전국최우해적철포。재증강금기수색배절，본문제출료일충기우위철함수적후선해린역구조방법。동시，급출료본문산법이개솔1수렴우전국최우적이론증명。대허의장경화진실장경적실험결과표명본문산법가이쾌속획취섭상궤구진고계적전국최우해。
Resection is one of important issues in machine vision.Although L2 norm based least square method is reasonably fast,the globally optimal solution cannot be obtained theoretically due to its non-convexity of the objective function .Optimization using the L∞norm has been becoming an effective way to solve parameter estimation problems in multiview geometry .But the computational cost increases rapidly with the size of measurement data .In the paper,we propose a novel approach under the frame-work of enhanced continuous taboo search (ECTS)for resection in multiview geometry .ECTS is an optimization method in the do-main of artificial intelligence,which has an interesting ability of covering a wide solution space by promoting the search far away from current solution and consecutively decreasing the possibility of trapping in the local minima .We propose the corresponding ways in the key steps of ECTS,diversification and intensification .We also present theoretical proof to guarantee the global conver-gence of search with probability one .Experimental results validate that the ECTS can obtain the global optimum effectively and effi-ciently .Potentially,the novel ECTS framework can be employed in many applications of multi-view geometry .