3D Radar Object Detection Using Sub-Image Extraction and Model Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection systems for radar images take a long time to detect objects, and they struggle to reduce processing time without compromising detection accuracy when using learned models.
Innovation Solution
A detection apparatus that includes a position determination unit, an extraction unit, a model selection unit, and a detection unit. This apparatus extracts a 3D sub-image from a 3D radar image based on the subject's position and type, selects a learned model based on the sub-image size and type, and uses the model to detect objects in the sub-image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a learned model is used for object detection in radar images, then detection accuracy is improved, but processing time increases
Solution Approach 1:
The patent divides the large 3D radar image into multiple smaller 3D sub-images based on detected subject positions. By processing only relevant sub-regions instead of the entire image, the computational load on the learned model is reduced while maintaining detection accuracy for objects of interest.
Solution Approach 2:
The patent extracts only the necessary portions (3D sub-images) containing potential objects from the complete radar image. This extraction is guided by subject position information, allowing the system to focus computational resources on regions where objects are likely to be found, thereby reducing processing time without compromising detection accuracy.
2Reliability
If the entire 3D radar image is processed, then detection coverage is improved, but processing time increases
Solution Approach 1:
The patent segments the 3D radar image into multiple 3D sub-images based on subject positions. This segmentation allows the system to maintain comprehensive detection coverage by processing only relevant regions containing subjects, rather than processing the entire image, thus reducing processing time while preserving detection reliability.
Solution Approach 2:
The patent performs preliminary detection of subject positions before extracting 3D sub-images for detailed object detection. This preliminary action enables the system to identify and focus on regions containing subjects, ensuring comprehensive coverage of potential objects while avoiding unnecessary processing of empty regions, thereby reducing overall processing time.
3Measurement precision
If multiple learned models are used for different subject types, then detection accuracy is improved, but device complexity increases
Solution Approach 1:
The patent applies different learned models selectively based on the type of subject detected in each 3D sub-image. By matching specific models to specific subject types (e.g., different models for different object categories), the system achieves high detection accuracy for each class while managing complexity through organized model selection based on subject classification.
Solution Approach 2:
The patent changes the parameter of model selection based on subject type classification. Instead of using a single model for all cases, the system dynamically selects appropriate models based on the detected subject category, thereby improving detection accuracy for different object types while maintaining manageable system complexity through parameter-based model selection.
Data Source
AI summary
The detection apparatus (100) of the first example embodiment includes a position determination unit (12), an extraction unit (14), a model selection unit (16), and a detection unit (18). The position determination unit (12) determines a position of a subject in a 3D radar image. The extraction unit (14) extracts a 3D sub-image from the 3D radar image by using a reference position based on the determined position of the subject and one of extraction sizes specified for each type of subject. The model selection unit (16) selects at least a learned model based on at least one of a size of the 3D sub-image and a type of the subject included in the 3D radar image. The detection unit (18) which detects an object in the 3D sub-image by using the selected learned model.