Abstract: An image mosaic algorithm based on machine vision software MVTec Halcon is proposed.Firstly,the camera calibration,as well as images preprocessing are performed.Secondly,feature points are extracted by Harris corner detection and then the points are matched using normalized cross-correlation.After this,error matching points are excluded using RANSAC (random sample consensus) and homography matrix is estimated.At last,images are fused by the method of improved weighted fusion.The software correspond to the algorithm which were proposed in this paper has been worked out and mosaic experimental has been finished.Experimental results show that image mosaic can be accomplished rapidly and efficiently by this method and precision is 0.02 mm.