3D object detection method based on monocular camera

Abstract

The invention relates to a 3D object detection method based on a monocular camera, and the method comprises the steps: building a depth estimation model which is used for obtaining a prediction depth map matched with original image data; obtaining original image data through a vehicle-mounted camera; acquiring a predicted depth map matching the original image data by using the depth estimation model; detecting a target object in the original image data; and projecting the target object to the corresponding predicted depth map to generate an anchoring area, and performing 3D reconstruction on the anchoring area to obtain a three-dimensional coordinate value of the target object in the world coordinate system. According to the 3D object detection method, the three-dimensional coordinate information of the object can be obtained only by means of the monocular camera, the invention does not depend on the assumed basis that the road surface is completely flat, the cost is low, the detection precision is high, accurate reference data can be provided for a driver, the driving safety can be improved, and the invention has important use value.

Description

technical field [0001] The invention relates to the technical field of 3D object detection, in particular to a 3D object detection method based on a monocular camera. Background technique [0002] Detecting objects of interest and inferring their 3D properties is a central problem in computer vision and has achieved widespread applications. Especially in the past ten years, with the rapid development of unmanned driving technology and mobile robots, object detection has played an extremely important role in the perception system. Accurate and efficient perception systems can effectively ensure that the robot and other surrounding moving objects safety. In recent years, although 2D object detection has also been rapidly developed in unmanned driving systems, more improvements are still needed to convert detected objects from the image plane to real-world poses. The task of conventional 3D object detection usually relies heavily on depth sensors such as lidar or millimeter-w...

AI Technical Summary

Problems solved by technology

[0004] However, the actual road surface cannot be completely flat. When the road surface is curved or uneven, these traditional methods will be affected.

For example, when the ground plane is assumed to be flat, it is actually uneven, curves on the driving surface can lead to inaccurate predictions, and distance estimates to obstacles in the environment can be over- or under-judged

In both cases, inaccurate distance estimation can have a direct negative impact on various maneuvers of the vehicle, potentially compromising lateral and longitudinal control or drivin

Images