In the real scene,the viewing angle range of the image obtained by the camera is often limited,and the actual demand requires the panorama of the scene.Aiming at the demand for panoramic images in daily life and industrial production.The traditional random sample consensus (RANSAC) algorithm has the defects of mismatching point pairs and low r egistration speed in image registration because there is no upper limit on the number of iteration s.This paper proposed an improved RANSAC algorithm to improve the efficiency of panoramic image stitching.The improved RANSAC algorithm finds an optimal data detection model by detecting points in the circle,and continuously updates the coordinates of the iterative circle center through the particle swarm algorithm.Finally,an optimal matching model is obtained to eliminate the outliers in the feature point matching process.It can improve the accuracy of feature point registration and reduce the complexity of the algorith m.The experiment of multi group image stitching shows that the average correct matching rate of t he improved RANSAC algorithm proposed in this paper is 9.0575% higher tha n that of other algorithms,and the average registration rate of the algorithm is 5.17375s h igher,which realizes a more robust panoramic image stitching effect.