AI intelligent object recognition method based on 3D vision

Abstract

The invention discloses an AI intelligent object recognition method based on 3D vision which comprises the steps that a virtual three-dimensional coordinate system is established, three fisheye cameras are symmetrically placed with the original point of the three-dimensional coordinate system as the center, and a checkerboard calibration plate is arranged in front of each fisheye camera; a plurality of angular points are selected from each checkerboard calibration plate to serve as feature points, three-dimensional coordinates of the feature points are obtained, two-dimensional coordinates are obtained through an angular point detection algorithm, and then the conversion relation of each fisheye camera from a three-dimensional space to a two-dimensional space is calculated through a PNP algorithm, namely a rotation matrix and a translation matrix; the corresponding relation between each camera image and the three-dimensional space is obtained through the internal reference, the rotation matrix and the translation matrix of the fisheye cameras, and then the images collected by the three fisheye cameras are spliced into a 3D visual image; and constructing and training of a 3D visual image recognition network are carried out, inputting of the 3D visual image is carried out, and a recognized object is displayed on the image. The invention belongs to the technical field of information, and can obtain a scene image in a 360-degree view field range and accurately identify an object in a scene.

Description

technical field [0001] The invention relates to a 3D vision-based AI intelligent object recognition method, which belongs to the field of information technology. Background technique [0002] Object recognition is a method of judging whether there is a specific object in the image through computer vision technology or AI-based deep learning, including pedestrians, vehicles, traffic signs, animals, etc., and simultaneously calculates the orientation or distance of the object. This technology can be combined with object re-identification, object tracking and other technologies and applied to artificial intelligence systems, vehicle driving assistance systems, intelligent robots, human behavior analysis, intelligent transportation and intelligent monitoring and other scenarios. The research on object recognition system started in the mid-1990s. From the very beginning to 2002, researchers borrowed and introduced some mature methods in the field of image processing and pattern ...

AI Technical Summary

Problems solved by technology

In many scenes, all-round detection is required, and a single camera has a limited field of view and cannot cover all the field of view. Therefore, the traditional method is to place multiple cameras in the field of view for simultaneous recognition, but this method not only increases the system overhead, and it is not convenient to observe the test results

Images