Vehicle Exterior Recognition With Mixed Object Speed Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle exterior environment recognition systems face challenges in stably deriving the speed of three-dimensional objects, particularly when raindrops on image-capturing units lead to erroneous parallax information and abrupt speed changes, potentially causing unintentional acceleration and unsafe proximity to preceding vehicles.
Innovation Solution
A vehicle exterior environment recognition apparatus that combines stereo speed, monocular speed, and predicted speed using a mixture ratio, derived from luminance images and distance images, to stabilize the object speed derivation, incorporating a stereo weight, monocular weight, and prediction weight calculation to mitigate errors caused by raindrops or blurred images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If stereo speed derivation is used to obtain accurate object speed, then measurement precision is improved, but reliability deteriorates when raindrops or blurred images cause erroneous parallax information
Solution Approach 1:
The patent combines three different speed derivation methods (stereo speed from distance image, monocular speed from luminance image, and predicted speed from temporal changes) into a unified speed derivation system. By merging multiple independent speed estimation approaches, the system achieves both high measurement precision and reliability, as the combination compensates for the weaknesses of individual methods under adverse conditions like raindrops or image blur.
Solution Approach 2:
The patent dynamically changes the weighting parameters (mixture ratio) of different speed derivation methods based on image quality assessment. When raindrops or blur are detected, the system adjusts the parameters to reduce the weight of stereo speed derivation and increase the weight of monocular and predicted speed derivation, thereby maintaining reliable speed estimation despite degraded image quality.
2Reliability
If multiple speed derivation methods are combined to improve reliability, then device complexity increases due to multiple processing units
Solution Approach 1:
The patent implements a multi-functional speed derivation system where a single processing apparatus performs multiple functions: deriving stereo speed from distance images, deriving monocular speed from luminance images, calculating predicted speed from temporal changes, assessing image quality, and dynamically weighting the results. This universal approach consolidates what could be separate systems into one integrated unit, reducing overall device complexity while maintaining high reliability through the combination of multiple methods.
Data Source
AI summary
A vehicle exterior environment recognition apparatus includes a stereo speed derivation unit, a monocular speed derivation unit, a predicted speed derivation unit, a mixture ratio derivation unit, and an object speed derivation unit. The stereo speed derivation unit derives a stereo speed indicating a speed of a three-dimensional object extracted from a distance image derived from luminance images of two image-capturing units. The monocular speed derivation unit derives a monocular speed indicating a speed of the three-dimensional object extracted from the luminance image of one of the image-capturing units. The predicted speed derivation unit derives a predicted speed based on a past speed of the three-dimensional object. The mixture ratio derivation unit derives a mixture ratio of the stereo, monocular, and predicted speeds. The object speed derivation unit mixes the stereo, monocular, and predicted speeds at the mixture ratio, to derive an object speed.