The provided method includes: classifying damage assessing images of a same vehicle portion into one same image set; obtaining keypoint features of each image set, diving every two damage assessing images in each image set into one image group, matching multiple relevant keypoints from the damage assessing images in each image group; calculating a feature point transform matrix of each image, and converting one of the two damage assessing images in each image group to a to-be-verified image which has the same shooting angle as the other damage assessing image in the image group; matching feature parameters of the to-be-verified image with those of the other damage assessing image in the same image group; and generating reminder information, when there are unmatched parameters, to remind the user of frauds of the damage assessing images received from the terminal.