Monocular 3D detection method and system based on ground fitting and storage medium

By segmenting the ground point cloud and fitting the ground plane, optimizing the distance of the 3D detection frame, combined with the monocular image detection results, the deviation and real-time problems caused by insufficient depth information and model complexity in traditional 3D detection are solved, and more accurate and fast 3D object detection.

CN119027913APending Publication Date: 2024-11-26COWA TECHNOLOGY CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411114697.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-14
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

Traditional 3D detection based on monocular images causes the deviation of the 3D detection frame in 3D space due to the lack of depth information, and the 3D detection model based on radar point cloud is too complex and difficult to meet the real-time requirements.

Method used

By segmenting the ground point cloud and fitting the ground plane, the distance of the 3D detection frame is optimized and combined with the monocular image detection results are optimized to reduce detection errors and improve real-time performance.

Benefits of technology

The accuracy and real-time performance of 3D detection results are improved, the detection speed is improved, the computing resource consumption is reduced, and the detection results are more in line with reality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119027913A_ABST
    Figure CN119027913A_ABST
Patent Text Reader

Abstract

The invention discloses a monocular 3D detection method and system based on ground fitting and a storage medium, and the method comprises the following steps: S1, segmenting a ground point cloud and fitting a ground plane based on radar point cloud data; s2, obtaining a 2D target detection frame result and a 3D target detection frame result based on the monocular image data; and S3, using the 2D target detection frame obtained in the step S2 and the ground fitting result obtained in the step S1 to optimize the 3D target detection frame result in sequence so as to obtain the optimized 3D target detection frame result. According to the invention, the ground plane fitting result and the monocular 3D detection result are combined, and the 3D detection result is corrected by using the ground fitting result, so that the detection precision of the monocular 3D detection result is improved.
Need to check novelty before this filing date? Find Prior Art