Improved Genetic Algorithm for Bundle Adjustment in Photogrammetry

In this work, a Constraint Law method (CLM) based Genetic Algorithm (CLM-GA) is proposed for Nonlinear Least Square (LS) Regression and Normal Optimization Problems (NOP). Numerical experimental results show that CLM-GA is more robust than traditional methods (Gauss Newton, Levernberg Marquardt, EM-GA, EDBF-GA) for both LS and NOP problems in three aspects: 1) efficiency; 2) accuracy; 3) astringency. There are many LS and NOP applications in the physical world, such as military, economics, industry, photogrammetry and so on. Research of this paper can be easily implemented and applied for those applications.