Method for generating sample images under all vision angles on basis of standard image

Abstract

The invention relates to the technical field of pattern recognition, in particular to a method for generating sample images under all vision angles on the basis of a standard image. The method comprises the following steps: firstly, a plurality of parameters are adopted to determine the space three-dimensional positions where a camera is located when in shooting, all values of the parameters are combined to be capable of covering all the positions where the camera can be located in the three-dimensional space, and each group of values only corresponds to a certain position in the space; then two groups of different parameters corresponding to two different space positions of the camera are used for showing the transformation relation among real images shot by the camera at the two positions; and then the transformation relation is applied to the real images shot at the first camera position, and a sample image can be generated. The image transformation relation leads the generated sample image to be capable of simulating the real images shot from corresponding vision angles completely, and can generate needed image under specified vision angle by controlling the values of the parameters.

Description

technical field [0001] The invention relates to the technical field of pattern recognition, in particular to a method for generating sample images at various viewing angles based on a standard image. Background technique [0002] Sample image generation is an image transformation method, which is an important part of the supervised learning method in the field of pattern recognition and machine learning. The purpose of sample image generation is to perform a series of image transformations on one or several standard images (usually images taken from the front of the target object) taken in real conditions, and then generate various samples. These sample images have a certain degree of similarity with the real images obtained by the camera shooting the target object under various viewing angles, so they can be used to simulate the real images taken. Effective sample image generation methods can generate sample images that are highly similar to real images. The supervised lea...

AI Technical Summary

Problems solved by technology

[0011] In view of this, the main purpose of the present invention is to provide a sample image generation method for the supervised learning process in the field of pattern recognition and machine learning. The sample image generated by this method can completely simulate the real image taken from the corresponding perspective, and can Provide more accurate prior knowledge reserves for pattern recognition and machine learning, and then obtain a higher success rate of object recognition, which can solve problems such as realizing the prior knowledge reserve of target objects to successfully identify target objects in a short period of time in the battlefield. problem, so as to provide a method for generating sample images under various viewing angles based on a standard image, this method overcomes the problem that the transformation matrix used in the process of generating sample images in existing methods cannot fully represent the transformation relationship between two images defect, which improves the similarity between the generated sample image and the real image

Images