3D Scene Object Detection via Floor Plane Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image processing technologies fail to effectively estimate object locations from stereoscopic pairs of images, particularly in complex scenes with occlusions and noise, and struggle to reconstruct amorphous objects like flames and smoke.
Innovation Solution
A method that converts a stereoscopic pair of 2-D images into a collection of 3-D objects by computing depth maps, estimating camera extrinsic parameters, and projecting 3-D objects onto a plane parallel to the floor, allowing for the identification of objects through symmetry fitting and peak detection in blurred images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional image processing methods are used to estimate object locations from stereoscopic pairs, then the processing is simpler, but the accuracy is insufficient especially in complex scenes with occlusions and noise
Solution Approach 1:
The patent segments the image processing task into multiple distinct stages: depth map computation from stereoscopic pairs, floor plane estimation, 3-D point recovery, and object identification through symmetry fitting. This segmentation allows each stage to be optimized independently, improving overall accuracy while managing complexity through modular processing
Solution Approach 2:
The patent transitions from 2-D image processing to 3-D scene reconstruction by computing depth maps and recovering 3-D points. This dimensional transformation enables more accurate object location estimation by utilizing depth information, while the structured 3-D representation helps manage the complexity of processing occluded and noisy data
2Reliability
If 3-D scene reconstruction is performed to detect objects in occluded areas, then object detection capability is improved, but the processing time and computational load increase
Solution Approach 1:
The patent performs preliminary estimation of floor plane parameters and camera extrinsic parameters before object detection. This preliminary action establishes a reliable 3-D geometric framework that accelerates subsequent object detection in occluded areas, reducing overall processing time while maintaining detection reliability
Solution Approach 2:
The patent changes the parameter representation from 2-D image coordinates to 3-D spatial coordinates with explicit floor plane references. This parameter transformation enables more efficient detection of objects in occluded areas by utilizing the structured 3-D geometry, balancing detection reliability with processing efficiency
3Measurement precision
If symmetric figures are fitted to points in the projection to identify objects, then object identification accuracy is improved, but the method becomes less adaptable to asymmetric objects
Solution Approach 1:
The patent acknowledges the limitation of symmetric figure fitting for asymmetric objects and incorporates this understanding in the object identification process. The method uses symmetric figures as a starting point but allows for asymmetric variations, enabling the system to maintain good identification accuracy for symmetric objects while remaining adaptable to asymmetric ones through the flexible 3-D point cloud representation
Data Source
AI summary
Systems and methods for processing a pair of 2-D images are described. In one example, a stereoscopic set of images is converted into a collection of regions that represent individual 3-D objects in the pair of images. In one embodiment, the system recovers the 3-D point P for each point p that appears in both images. It estimates the 3-D orientation of the floor plane, and the image capture planes and their height from the floor. The system then identifies the collection B of points P that do not represent points on the floor and generates a projection C of B onto a plane parallel to the floor. It blurs the projection C and identifies peaks in the blurred image, then fits symmetric figures to the points in C around the identified peaks. The system projects the 3-D figures associated with the symmetric figures back onto the 2-D images.