用于胶囊内窥镜方位求解的LM人工蜂群算法

An advanced algorithm consisting of Levenberg-Marquardt and Artificial Bee Colony Algorithm applied to locate capsule endoscope

目的:为了解决胶囊内窥镜的定位问题,为活检和喷药提供方位信息,提出了反馈式交流励磁跟踪方法。基于Levenberg-Marquardt算法对初值的依赖和局部优化能力以及人工蜂群算法的随机性和全局优化能力,提出了一种改进的人工蜂群算法和Levenberg-Marquardt算法相结合的新型混合全局算法来解决定位问题。方法:根据方位跟踪原理建立跟踪模型,然后用此算法求解跟踪数学模型,求得胶囊的位置和方向信息。采用Visual Studio 2010进行了算法仿真实验,本文选择根据实际情况从螺旋线上有规律的选取100个点,并将100个(x,y,z,α,β)方位坐标的代入模型公式得到对应的100组电压值,然后再利用本文提出的混合算法求得100组电压值对应的100组方位值。以此方法来验证算法的正确性和高效性。结果:实验结果表明该定位算法能成功求解磁跟踪非线性方程组,且位置和方向误差均极小,迭代次数也在允许范围之内。结论:此算法充分利用了LM的局部寻优能力和ABC的全局寻优能力,不但克服了LM对初始向量选择的依赖,补偿了ABC算法定位效率不高的缺点,而且算法的求解精度高,收敛速度快,具有很好寻优能力。

Objective To solve the problem of locating the capsule endoscope, and to provide location information for biopsying and spraying, a non-contact tracking method on electromagnetic induction is proposed. Based on that Levenberg- Marquardt （LM） Algorithm, depending on the initial value, has a strong competence of local optimization, and that the Artificial Bee Colony （ABC） Algorithm has the characteristics of randomness and the competence of global optimization, an hybrid algorithm consisting of the LM Algorithm and ABC Algorithm is advanced. Methods The tracking model, based on the principles of position tracking, found the solution of the tracking mathematical model to obtain the position and location information of the capsule endoscope.Visual Studio 2010 was applied to conduct simulation of the algorithm,and 100 points from the helix were regularly selected based on the actual situation, and the 100 groups （x,y,z,α,β） of position were coordinated into the model equations to obtain 100 groups of corresponding voltage values. The 100 groups of voltage values obtained by the hybrid algorithm were compared to the 100 groups of position values to verify the correctness and efficiency of the algorithm. Results Experimental results showed that the algorithm successfully solved nonlinear equations of magnetic tracking model. The location and orientation errors were minimal, and the number of iterations was within the allowed limits. Conclusion This algorithm takes full advantage of the local searchingability of LM Algorithm and the global optimization ability of ABC Algorithm.Moreover, it not only overcomes the dependence on the initial vector of LM Algorithm, but also compensates for the inefficient locating of ABC Algorithm. This algorithm has a high-precision for solving, a fast convergence, and a good optimization ability.