Three-dimensional target recognition method and device in image, storage medium and equipment

Abstract

The application discloses a three-dimensional target recognition method and device in an image, a storage medium and equipment, and belongs to the technical field of unmanned driving. The method comprises the following steps: acquiring multiple ring-view images, wherein the multiple ring-view images are composed of images obtained by simultaneously photographing a same scene by multiple cameras with different horizontal photographing angles in a vehicle; performing feature extraction on each ring-view image to obtain multiple feature maps; performing two-dimensional target recognition on the multiple ring-view images to obtain the type, position and size of a detection frame; performing three-dimensional feature processing on the type, position and size of the detection frame respectively, splicing processing results to obtain k three-dimensional query features; and processing the k three-dimensional query features and the multiple feature maps to obtain a three-dimensional target recognition result. The application can utilize two-dimensional and three-dimensional related information to enhance the three-dimensional target recognition result, can simultaneously output the two-dimensional target recognition result and the three-dimensional target recognition result, and the two recognition results are self-consistent, thereby facilitating post-processing fusion.

Images