Road Object Detection Using NURBS 3D Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for detecting and modeling road objects and paint on a surface fail to provide accurate three-dimensional positions and logical information, especially when dealing with complex shapes and connectivity between frames, leading to significant errors.
Innovation Solution
A method that involves scanning the road, generating a 3D model, creating a top-view image, detecting objects, projecting them onto the 3D model, and using Non-Uniform Rational B-Spline (NURBS) curve fitting for precise 3D modeling, while merging data from multiple vehicles to refine road surface estimation and object detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional satellite navigation systems are used to determine vehicle position, then the positioning can be achieved with standard equipment, but the accuracy is insufficient (cannot achieve a few centimeters accuracy)
Solution Approach 1:
The patent introduces road markings and road objects as intermediary reference elements between the vehicle and satellite navigation system. By detecting and tracking these intermediate features on the road surface, the system achieves high-precision positioning (few centimeters accuracy) that cannot be obtained through satellite navigation alone. The road markings serve as a mediator that bridges the gap between low-precision satellite data and high-precision vehicle position requirements.
2Manufacturing precision
If aerial photographs or satellite images are used to capture road markings, then a perpendicular view with little distortion is obtained, but sufficient detail for generating highly accurate maps is not provided
Solution Approach 1:
The patent merges multiple data sources including aerial photographs, satellite images, and ground-based camera captures to create a comprehensive road map. By combining the distortion-free perpendicular view from aerial/satellite imagery with the detailed ground-level perspective from vehicle-mounted cameras, the system achieves both high map accuracy and rich road feature details that neither source could provide alone.
Solution Approach 2:
The patent transitions from two-dimensional aerial/satellite imagery to three-dimensional road surface modeling by incorporating ground-based camera perspectives. This dimensional change allows the system to capture road markings and objects from multiple angles and depths, creating a rich 3D representation that preserves fine details while maintaining accurate geometric relationships.
3Productivity
If conventional detection methods detect objects from every camera frame, then object detection is performed continuously, but connectivity between detected results from different frames is very difficult to obtain
Solution Approach 1:
The patent implements feedback mechanisms where detected objects in one frame are used to guide and constrain detection in subsequent frames. By tracking object positions, shapes, and characteristics across frames and using previous detection results as feedback for current frame analysis, the system maintains object connectivity information while performing continuous high-frequency detection. This feedback loop ensures that detected objects form coherent trajectories rather than isolated detections.
4Ease of manufacture
If conventional methods represent detected objects with simple geometric shapes, then the modeling process is simple, but large errors occur since real-world objects have arbitrary shapes
Solution Approach 1:
The patent employs curved surface modeling techniques including splines and point cloud representations to accurately capture the arbitrary shapes of real-world road objects. Instead of forcing objects into simple geometric primitives, the system uses flexible curved mathematical models that can adapt to complex object geometries while maintaining computational efficiency. This approach preserves manufacturing simplicity through parametric modeling while achieving high shape accuracy.
Data Source
AI summary
A method for detecting and modelling of an object on a surface of a road by first scanning the road and generating a 3D model of the scanned road (which 3D model of the scanned road contains a description of a 3D surface of the road) and then creating a top-view image of the road. The object is detected on the surface of the road by evaluating the top-view image of the road. The detected object is projected on the surface of the road in the 3D model of the scanned road. The object projected on the surface of the road in the 3D model of the scanned road is modelled.