Obstacle Detection Using Predictive Velocity Vector Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional obstacle detection systems face challenges in stabilizing obstacle detection due to the motion of a moving vehicle, particularly in canceling out image shifts caused by pitch and yaw, which complicates the identification of obstacles in the vehicle's running direction.
Innovation Solution
An obstacle detection apparatus using an image sensor on a moving vehicle, which includes an image input unit, a velocity detector, a predictive velocity vector calculator, and a determination unit to identify obstacles by comparing the horizontal components of detected and predictive velocity vectors, thereby reducing the impact of vehicle motion on obstacle detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If optical flow analysis is used to detect obstacles, then obstacle detection capability is provided, but detection stability deteriorates due to vehicle motion causing varying optical flow
Solution Approach 1:
The patent extracts and removes the component of image shift caused by vehicle pitch and yaw motion from the overall image sequence. By identifying and eliminating this specific disturbance factor, the system isolates the true obstacle motion signal, thereby improving detection stability while maintaining the ability to handle varying vehicle conditions.
Solution Approach 2:
The patent introduces a reference object (such as the road surface or stationary background elements) as an intermediary to measure and compensate for vehicle-induced image shifts. This reference serves as a mediator between the vehicle motion and the obstacle detection process, allowing the system to distinguish between shifts caused by vehicle motion and those caused by actual obstacles.
2Reliability
If image shift cancellation techniques are applied to compensate for pitch and yaw, then some stability is improved, but complete cancellation is difficult due to various movable objects in the image
Solution Approach 1:
The patent segments the image analysis process into distinct components: identifying stationary reference regions (such as road surface), calculating shift parameters from these regions, and applying compensation separately from obstacle detection. This segmentation allows the system to handle different types of motion independently, improving effectiveness without excessive complexity.
Solution Approach 2:
The patent applies partial compensation by focusing on canceling the dominant shift components caused by pitch and yaw, rather than attempting to compensate for all possible image variations. By addressing the most significant sources of instability, the system achieves practical effectiveness while avoiding the complexity of complete compensation for all movable objects.
3Loss of information
If velocity vectors are calculated for all points in the image, then comprehensive motion information is obtained, but processing complexity increases
Solution Approach 1:
The patent extracts velocity information only from specific reference regions (such as the road surface or stationary background areas) rather than calculating velocity vectors for all image points. This selective extraction maintains sufficient motion information for obstacle detection while significantly reducing processing complexity.
Solution Approach 2:
The patent makes the calculated velocity vectors from reference regions serve multiple functions: they are used both to compensate for image shifts and to provide baseline motion information for obstacle detection. This multi-functionality reduces the need for separate calculations, thereby lowering overall processing complexity while maintaining information completeness.
Data Source
AI summary
An obstacle detection apparatus includes an image input unit inputting a image sequence captured by an image sensor, a velocity detector deriving velocity vectors representing motions of a to-be-detected point and a to-be-compared point on each image of the image sequence, a predictive velocity vector calculator calculating a predictive velocity vector of the to-be-detected point using a detected velocity vector of the to-be-compared point, an obstacle detector detecting the to-be-detected point as a point on the obstacle when a horizontal component of a detected velocity vector of the to-be-detected point is larger than a horizontal component of the predictive velocity vector by a given threshold, and an obstacle information output unit outputting information on the to-be-detected point detected as the obstacle.