Vehicle Rear Warning Apparatus Using Lane-Based Recognition Priority
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle rear warning systems issue false warnings due to misidentification of elongated objects, such as trees or parked vehicles, as pedestrians or two-wheeled vehicles, despite efforts to improve recognition accuracy.
Innovation Solution
A vehicle rear left and right side warning apparatus that uses pattern recognition technology to detect 3D objects and adjusts recognition priority based on the driving lane, setting lower priority for types of objects less likely to be present, thereby reducing false warnings by tightening the threshold for shape matching.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pattern recognition technology is used to detect 3D objects in images captured by side cameras, then the system can identify objects such as vehicles and pedestrians, but false warnings are issued when elongated objects like trees or parked vehicles are misidentified as warning targets
Solution Approach 1:
The patent applies local quality by setting different recognition priorities for different object types based on their likelihood of being present in specific driving lanes. For example, pedestrians have high recognition priority in areas near sidewalks, while parked vehicles have low priority in moving traffic lanes. This localized adjustment of recognition thresholds reduces false warnings while maintaining detection accuracy for actual hazards.
Solution Approach 2:
The system dynamically changes the recognition threshold parameter based on the detected driving lane and contextual information. When the vehicle is in a lane where certain objects are unlikely to be present (e.g., parked vehicles in a moving traffic lane), the recognition threshold for those object types is raised, making the system more selective and reducing false positives while maintaining sensitivity for relevant objects.
2Productivity
If the recognition threshold is lowered to detect all potential objects, then more objects are detected, but the system cannot distinguish between actual warning targets and non-targets such as planted trees
Solution Approach 1:
Different recognition priorities are assigned to different object types based on their contextual relevance to driving safety. Pedestrians and moving vehicles maintain high recognition priority across most areas, while stationary objects like parked vehicles and planted trees have context-dependent priorities. This localized quality adjustment allows comprehensive detection of potential hazards while filtering out non-threatening objects.
Solution Approach 2:
The patent introduces an intermediate filtering mechanism that uses driving lane information and object characteristics as mediators between raw detection and warning issuance. The system first detects all objects, then applies intermediate filtering based on whether the object type is appropriate for the current driving context, and only issues warnings for objects that pass this intermediate evaluation stage.
3Device complexity
If simple pattern recognition is used to detect objects, then the system is computationally efficient, but it misidentifies elongated objects such as planted trees as two-wheeled vehicles or pedestrians
Solution Approach 1:
The system applies local quality by using simple pattern recognition for initial detection but adds context-specific verification only where needed. For example, elongated objects detected in areas where trees are unlikely (based on driving lane and location data) undergo additional classification checks, while objects in high-risk areas receive immediate attention. This selective application of complex analysis maintains efficiency while improving accuracy for ambiguous cases.
Solution Approach 2:
The system performs preliminary filtering using simple pattern recognition to quickly identify potential objects, then applies more sophisticated classification only to objects that require further verification. This preliminary action separates the majority of clear-cut cases (handled by simple recognition) from ambiguous cases (requiring additional analysis), maintaining computational efficiency while improving classification accuracy for problematic objects.
Data Source
AI summary
A 3D object detecting device includes a driving-lane detecting unit that detects a driving lane of a vehicle; a 3D-object detecting unit that detects, by using pattern recognition technology, a type of 3D object with respect to images captured by side cameras; and a recognition-priority setting unit that varies and sets a recognition priority for the detected type of 3D object, based on the detected driving lane. In accordance with the driving lane of the vehicle, the recognition-priority setting unit sets the priority for pattern recognition lower for a type of 3D object that is less likely to exist in rear left and right sides of the vehicle, to thereby reduce the possibility that a 3D object that should not be regarded as a warning target is erroneously detected from the images as a warning subject.