In order to solve the problem that images shot under industrial environment is considerably influenced by noise and distortion, precision of traditional algorithms for calibrating matching parameters cannot meet field measurement demand, a new method for calibrating large field of view image matching parameters based on non-metric correction is proposed. A layout plan of colinear feature points is designed, and based on non-metric correction, image coordinates of colinear feature points can be undistorted, which can be recognized and matched automatically with the aid of layout plan. Modified random sampling consensus algorithm based on regional division is proposed, and matching parameters can be calibrated. Experimental results show that matching precision is increased by 51% at least, which meets precision requirements under industrial environment.