A Fast Registration Method Applied to Industrial Images

Abstract

The invention discloses a rapid registration method applied to industrial images, which includes the following steps: S100: select a specific area in a standard image, and manually set a corresponding reference ROI for the specific area; S200: set a scaling factor, and manually set Scale the standard image with reference to the ROI to obtain the first zoom image, and at this time the reference ROI in the standard image is also scaled accordingly; apply the canny operator to extract the edge points of the first zoom image; S300: Do a matching search on the test image ; S400: Set the same zoom factor as in step S200 to zoom the test chart to obtain a second zoom map, and confirm the preliminary matching position in the second zoom map based on the matching search range when doing the matching search in step S300; S500: Perform fine correction on the matching position obtained in step S400 to obtain the best matching position; S600: establish a matching coordinate system according to the best matching position in step S500, and perform matching at other positions in the test map.

Description

technical field [0001] The invention relates to the technical field of image registration, in particular to a rapid registration method applied to industrial images. Background technique [0002] The purpose of image registration technology is to find the corresponding position in the standard image in the target image, which is widely used in industrial vision. Currently, image registration algorithms on the market are mainly based on two directions: template matching, template matching (Blocking Matching) ) is based on the known template image to find a sub-image similar to the template image in another image. This technology can only be applied to images with small image rotation and local stretching. It can be more accurate on such images. The matching effect is relatively high, and the versatility is relatively high; however, in the case of image rotation and stretching, more accurate matching cannot be achieved, the flexibility of template matching is relatively poor a...

AI Technical Summary

Problems solved by technology

[0002] The purpose of image registration technology is to find the corresponding position in the standard image in the target image, which is widely used in industrial vision. Currently, image registration algorithms on the market are mainly based on two directions: template matching, template matching (Blocking Matching) ) is based on the known template image to find a sub-image similar to the template image in another image. This technology can only be applied to images with small image rotation and local stretching. It can be more accurate on such images. The matching effect is relatively high, and the versatility is relatively high; however, in the case of image rotation and stretching, more accurate matching cannot be achieved, the flexibility of template matching is relatively poor and the search range is too large, resulting in a long time-consuming matching; based on feature matching, the implementation Feature matching first extracts the features of the image, then generates the feature descriptor, and finally matches the features of the two images according to the similarity of the descriptor. The features of the image can be mainly divided into points, lines (edges), and regions (surfaces). Such features can also be divided into local features and global features. Regional (face) feature extraction is cumbersome and time-consuming, so point features and edge features are mainly used. This technology can be applied to complex situations where images have rotation and local stretching. However, the process of extracting image features is often time-consuming, and the extracted features are often only applicable to images of specific scenes, and the extracted local features are easily disturbed by noise, resulting in insufficient matching robustness.

Images