the aim of projective reconstructing is to estimate the position and direction of cameras through matching points in different images so lay the foundation for further reconstructing . on the basis of current methods of projective reconstructing, we used the levenberg-marquardt method to optimize the result of linear method so the precision is be improved, and we use the bundle adjustment method to entirely optimize the structure of scene and projective matrixes 本文在研究已有射影重建算法的基礎(chǔ)上,利用l-m算法對基于基礎(chǔ)矩陣的射影重建算法得到的線性結(jié)果進行優(yōu)化,提高了算法的估計精度和穩(wěn)定性,并在求得所有圖象對應(yīng)的投影矩陣后利用bundleadjustment方法對空間結(jié)構(gòu)及投影矩陣進行全局優(yōu)化。