网格搜索法优化的支持向量机室内可见光定位

采用信号强度特征建立指纹库,通过网格搜索法对支持向量机参数进行优化,利用SVM回归算法建模位置坐标和信号强度特征的映射关系,实现对待定位点位置坐标的预测。在待定位点误差范围内建立子集指纹库,根据欧式距离的远近分配权值,对预测到的坐标进一步优化,实现误差最小化。将没有优化的支持向量机与用网格搜索法、蚁群算法、粒子群算法优化后进行对比,实验结果表明,使用网格搜索法优化后的SVM回归算法可以实现良好的定位效果,最终平均定位误差可达到0.042 m,且算法所需时间优于蚁群算法,寻找全局最优解优于粒子群算法。

Support vector machine indoor visible light positioning optimized by grid search method

Use signal strength characteristics to establish Fingerprint Database,and Method through Grid Search to optimize support vector machine parameters.Use the SVM regression algorithm to model the mapping relationship between position coordinates and signal strength characteristics,and realize the positioning points'position coordinates.Establishing a subset fingerprint library within the error range of the points to be located,assign weights according to the distance of the Euclidean distance,and further optimize the predicted coordinates to minimize the error.Compared with the optimized Grid Search Method,Ant Colony Optimization,and Particle Swarm Optimization,the unoptimized support vector machine is compared.The experimental results show that the SVM regression algorithm optimized by the grid search method can achieve good positioning results.The final average positioning error can reach 0.042 m.Besides,the algorithm's time is better than that of the ACO,and the search for the optimal global solution is better than the PSO.