Image body identification, correction and registration method

Abstract

The invention discloses an image body identification, correction and registration method comprising the following steps: target detection: performing threshold segmentation by observing the features of an image, and roughly distinguishing between the foreground and the background; binarization: getting a black-and-white binary image, removing noise points through expansion, and retaining a compartment part in the original image; image feature extraction: positioning the four corners of the compartment based on the mathematical properties of trapezoid; perspective transformation: stretching an irregular convex quadrilateral to a regular rectangle through a transformation matrix; and registration and splicing: combining multiple box body images, and for the images with common image features, registering and splicing multiple box body images on the side by use of an Opencv open-source library Stitcher class or augmenting the images based on the properties of the matrix. The image body identification, correction and registration method is beneficial to eliminating background interference, reducing the sample set of box body identification and improving the efficiency of identification.

Description

technical field [0001] The invention belongs to image processing technology, in particular to an image subject recognition, correction and registration method based on image features and pixel distribution patterns. Background technique [0002] With the continuous improvement of the knowledge system of mathematics, physics, and geography, and the advancement of satellite technology, computer technology, and Internet technology, there are more and more image data, and the technology involved in image processing, storage, and retrieval is big data. A weapon of the times. Image subject recognition, correction and registration methods are widely used in fields such as geography and navigation with a large amount of scattered image data. In the field of transportation, the box identification of port containers can effectively improve transportation efficiency, reduce costs, and have great economic benefits. The recognition, correction and registration methods of the image subj...

AI Technical Summary

Problems solved by technology

[0004] At present, in the industrial field, pattern recognition is basically based on the original image for direct enhancement, clustering, and recognition. However, for images with a small proportion of the target foreground, it cannot effectively identify and distinguish the foreground and background, which increases the error in the processing and operation process and reduces the Accuracy and success rate of image analysis

Images