摘要
针对根系探地雷达数据重构中计算复杂和准确率低的问题,提出粒子群(PSO)与模拟退火(SA)相结合的正交匹配追踪(OMP)优化算法。首先,由Gabor原子对信号进行稀疏表示,建立求解空间;然后,以匹配函数为适应度,通过PSO算法求出匹配函数的最佳适应度值,找出可行解原子,再利用SA算法对PSO极值进行退温搜索,得出全局最优原子;最后,利用最优原子完成根系稀疏数据的重构。对A-scan数据和B-scan数据进行实验,结果表明,PSO-SA-OMP算法比传统OMP算法的计算时间减少了10.471 s和20.260 s,均方误差减小了1.225,信噪比提高了5.539 d B。
In view of the computational complexity and low accuracy on radar data reconstruction,we proposed an orthogonal matching pursuit（ OMP） based on particle swarm algorithm（ PSO） and simulated annealing（ SA）. We established a solution space by using Gabor,and calculated the fitness value of matching function by PSO. Then we got the best particle,and used SA algorithm for temperature annealing operation. Finally,we completed the sparse signal reconstruction by using the size and location of the optimal particle. By the experiments on A-scan data and B-scan data,PSO-SA-OMP was better than traditional OMP,with the computation time reduced by 10.471 s and 20.260 s,MSE reduced by 1.225,and PSNR improved by 5.539 dB,respectively.
出处
《东北林业大学学报》
CAS
CSCD
北大核心
2015年第8期120-124,共5页
Journal of Northeast Forestry University
基金
中央高校基本科研业务费专项资金(DL13BB21
DL13CB02)
黑龙江省自然科学基金项目(C201405)
黑龙江省博士后科研启动项目(LBH-Q14014)
