3D Object Detection Using Pseudo-3D Keypoint Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current 3D object detection methods in smart driving technologies face challenges in achieving accurate and efficient detection of objects in real-time, particularly in scenarios where high-cost hardware like depth cameras are not feasible, and there is a need for methods that can effectively utilize limited computing resources.
Innovation Solution
A method that involves obtaining 2D coordinates of key points in an image, constructing a pseudo 3D detection body in a 2D space, and determining the 3D detection body using depth information, allowing for accurate 3D object detection without relying heavily on neural networks or expensive hardware.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional 3D object detection methods using depth cameras or complex neural networks are employed, then detection accuracy can be improved, but hardware cost and computational resource consumption increase significantly
Solution Approach 1:
The patent creates a pseudo-3D detection body by copying and transforming the 2D detection body through geometric transformations. The 2D detection body is first transformed to a bird's-eye view coordinate system, then a pseudo-3D structure is constructed by adding depth information through scaling operations, avoiding the need for expensive depth cameras while achieving 3D detection capability
Solution Approach 2:
The patent replaces complex mechanical/optical systems (depth cameras, LiDAR) with computational geometry methods. Instead of using physical depth-sensing hardware, the system uses 2D image coordinates combined with geometric transformations and scaling to infer 3D spatial relationships, substituting mechanical depth sensing with mathematical computation
2Measurement precision
If high-precision 3D detection is achieved using existing methods, then object detection accuracy improves, but real-time processing capability deteriorates due to high computational requirements
Solution Approach 1:
The patent extracts only the essential 2D detection body from the image processing pipeline and separates the 3D reconstruction step. By taking out the 2D detection result and applying simple geometric transformations and scaling operations, the system achieves 3D detection without the heavy computational burden of full 3D neural network processing
Solution Approach 2:
The patent changes the parameter representation from complex 3D point cloud coordinates to scaled 2D coordinates with implicit depth information. By transforming the coordinate system and using scaling factors derived from object size information, the system maintains detection accuracy while reducing computational complexity for real-time processing
3Reliability
If conventional 3D detection approaches are used, then detection results can be obtained, but the method fails to effectively utilize limited computing resources in resource-constrained environments
Solution Approach 1:
The patent applies partial action by using only the necessary 2D detection body information and applying minimal geometric transformations to achieve 3D detection. Instead of processing full 3D point clouds or using complex neural networks, the system performs selective scaling and coordinate transformation only on the essential detection features, reducing energy consumption while maintaining reliability
Data Source
AI summary
A 3D object detection method includes: obtaining two-dimensional (2D) coordinates of at least one predetermined key point of a target object in an image to be processed; constructing a pseudo 3D detection body of the target object in a 2D space according to the 2D coordinates of the at least one predetermined key point; obtaining depth information of a plurality of vertices of the pseudo 3D detection body; and determining a 3D detection body of the target object in a 3D space according to the depth information of the plurality of vertices of the pseudo 3D detection body.