Feature vector extraction method, recognition method and corresponding device for 360-degree panorama picture

Abstract

The invention discloses a feature vector extraction method, a recognition method and a corresponding device of a 360-degree panorama picture. The method includes: obtaining a 360-degree panorama picture as a sample picture, and scaling the sample picture to a preset size W×H; extracting multiple sets of color difference vectors in two consecutive columns from the scaled sample picture to obtain multiple corresponding 360-degree The positive sample eigenvector of the panorama image; from the zoomed sample image, multiple sets of non-continuous two-column color difference vectors are obtained to obtain multiple negative sample eigenvectors corresponding to non-360-degree panorama images; for multiple positive sample eigenvectors and multiple Negative sample feature vectors are used for sample training to obtain support vector data for distinguishing features of 360-degree panoramic images and features of non-360-degree panoramic images. It can be seen that the present invention can obtain a large number of feature vectors of 360-degree panoramic pictures by using less resources of 360-degree panoramic pictures, thereby improving the recognition rate of 360-degree panoramic pictures.

Description

technical field [0001] The invention relates to the technical field of picture recognition, in particular to a feature vector extraction method, recognition method and corresponding device of a 360-degree panoramic picture. Background technique [0002] With the maturity of virtual reality technology (Virtual Reality, VR), resources applied in the VR mode are constantly emerging, especially various types of pictures suitable for viewing in the VR mode, such as 2D pictures, panoramic pictures, and the like. When watching various types of pictures in VR mode, the types of pictures are different, and the picture viewing tools are also different. In order to realize the automatic selection of image viewing tools instead of manual selection by the user, it is necessary to automatically identify the type of image, especially for the identification of 360-degree panoramic images. [0003] In the prior art, in the recognition method of a 360-degree panoramic picture, the features o...

AI Technical Summary

Problems solved by technology

[0004] In view of the relatively few 360-degree panorama resources in the prior art at this stage, it is relatively difficult to obtain a large number of picture samples, which will lead to the problem of difficulty in obtaining feature samples, a method for extracting feature vectors of 360-degree panorama pictures of the present invention, 360 A recognition method and a corresponding device for a panoramic picture, so as to solve or at least partly solve the above-mentioned problems

Images