GSA-SVM模型在GNSS高程拟合中的应用

Application of GSA-SVM Model in GNSS Elevation Fitting

在SVM模型的基础上,引入遗传算法与模拟退火算法,提出了一种GSA优化的SVM新模型。该模型实现参数寻优的主要思路为:首先利用GA全局搜索能力对SVM参数空间进行全局搜索,并将搜索得到最优解作为SA算法初始值;其次发挥SA算法在局部搜索中的优势,并将SA算法获得的解作为新一轮GA算法的初始值。通过上述全局搜索与局部搜索的不断迭代实现SVM模型最优参数的获取。使用某测区控制点GNSS观测数据对本文提出的高程拟合模型进行检验,结果表明,与传统的SVM模型相比,本文提出的GSA-SVM模型的高程拟合精度更高,在实际工程项目中的应用价值更高。

Based on the SVM model,this paper introduces genetic algorithm and simulated annealing algorithm,and proposes a new SVM model optimized by GSA.The main idea of parameter optimization of the model is as follows:firstly,the global search ability of GA is used to search the SVM parameter space,and the optimal solution is taken as the initial value of SA algorithm;secondly,it gives full play to the advantages of SA algorithm in local search,and takes the solution obtained by SA algorithm as the initial value of a new round of GA algorithm.Through the continuous iteration of global search and local search,the optimal parameters of SVM model are obtained.Using the GNSS observation data of the control point in a survey area to test the elevation fitting model proposed in this paper,the results show that compared with the traditional SVM model,the elevation fitting accuracy of the GSA-SVM model proposed in this paper is higher and has higher application value in practical engineering projects.