Vehicle Image Processing Resolving 3D Object Misrecognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current camera-based systems for sensing parking bays often misrecognize three-dimensional objects, such as a bumper or side sill, as the parking bay due to their appearance similar to white lines in the image, leading to incorrect identification.
Innovation Solution
A vehicle-mounted image processing device that includes an image acquisition unit, feature amount extraction unit, three-dimensional object information storage unit, and a feature amount distinguishing unit to differentiate between features on the road surface and three-dimensional objects by analyzing the positional relationship of line segments and using triangle calculations to determine if three-dimensional object information is present within a specific triangle formed by the feature points.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If edge information is used to sense parking bays based on brightness difference, then parking bay sensing capability is improved, but misrecognition of three-dimensional objects as parking bays occurs
Solution Approach 1:
The patent transitions from two-dimensional image plane analysis to three-dimensional spatial analysis by introducing height information. The processing unit calculates the height of edge-containing regions and compares it against a threshold to distinguish between parking bay markings (height ≈ 0) and three-dimensional objects (height > threshold), thereby resolving the misrecognition problem while maintaining sensing accuracy
Solution Approach 2:
The patent introduces height information as an intermediary parameter between edge detection and parking bay recognition. This intermediate measurement serves as a discriminative feature that mediates between the conflicting requirements of detecting all edges (including parking bays) while filtering out false positives from three-dimensional objects
2Reliability
If shadow regions are filtered using brightness change direction, then shadow misrecognition is prevented, but three-dimensional object misrecognition persists
Solution Approach 1:
The patent adds the height dimension to the analysis, moving beyond the two-dimensional brightness change direction method. By measuring the vertical extent of edge-containing regions and comparing against a height threshold, the system can distinguish three-dimensional objects from both parking bays and shadows, resolving the insufficient feature discrimination problem
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A vehicle-mounted image processing device 1000 acquires an image around a host vehicle 10, senses a feature amount of a target object present on a road surface from the image, and relative position coordinates with respect to the host vehicle 10, acquires information of a three-dimensional object around the host vehicle 10 and relative position coordinates of the three-dimensional object, distinguishes whether the feature amount is a feature amount of a feature on the road surface or a feature amount of a feature on the three-dimensional object by using a positional relationship between the feature amount and the three-dimensional object, and recognizes the target object by using a feature on the road surface. As a result, a vehicle-mounted image processing device, which is capable of avoiding misrecognition of a part of a three-dimensional object as a target object when sensing the target object present on a road surface such as a parking bay by using a camera image, is provided.