Vehicle Camera Ranging with Dynamic Fusion of Two Distance Estimates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating three-dimensional positions and depths from images captured by vehicle-mounted cameras suffer from inaccuracies and instability in object ranging, particularly in autonomous driving scenarios.
Innovation Solution
An object ranging apparatus that combines two distance estimation methods - depth estimation using deep learning and three-dimensional position estimation via motion parallax - and adjusts the weighting based on vehicle steering angle and acceleration to stabilize and improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single distance estimation method is used, then the device complexity is reduced, but the measurement precision and reliability of object ranging deteriorate
Solution Approach 1:
The patent combines two different distance estimation methods (depth estimation and three-dimensional position estimation) into a unified system. The combining unit integrates results from both methods to produce a final ranging result, leveraging the complementary strengths of each approach to improve overall measurement precision while maintaining reasonable system complexity.
Solution Approach 2:
The patent creates a composite estimation system that combines multiple estimation methodologies. By weighting and integrating results from different estimation approaches based on their respective reliability under various conditions, the system achieves superior measurement precision compared to using a single method.
2Ease of operation
If a single distance estimation method is used, then the system is simpler to operate, but the reliability of object ranging deteriorates under varying vehicle conditions
Solution Approach 1:
The patent implements a dynamic selection and weighting mechanism that adapts to varying vehicle conditions. The combining unit adjusts the weight given to each estimation method based on real-time vehicle state (steering angle, acceleration), ensuring reliable ranging results across different operating scenarios without requiring manual intervention.
Solution Approach 2:
The system uses feedback from vehicle state sensors (steering angle sensor, acceleration sensor) to dynamically adjust the weighting of different estimation methods. This closed-loop approach maintains high reliability by automatically compensating for conditions that may affect the accuracy of individual estimation methods.
3Ease of operation
If deep learning-based depth estimation is used alone, then the ease of operation is maintained, but the measurement precision deteriorates due to instability under vehicle motion
Solution Approach 1:
The combining unit acts as an intermediary that processes and integrates results from both depth estimation and three-dimensional position estimation. It applies weighting based on vehicle conditions to produce a final reliable distance measurement, effectively mediating between the simplicity of single-method operation and the precision of multi-method integration.
4Reliability
If three-dimensional position estimation using motion parallax is used alone, then the reliability improves, but the device complexity increases due to multiple image processing requirements
Solution Approach 1:
The combining unit serves multiple functions: it integrates results from both estimation methods, applies dynamic weighting based on vehicle conditions, and produces a final ranging result. This multi-functional component manages the complexity of processing multiple image-based estimations while maintaining system reliability.
Data Source
AI summary
An object recognition unit recognizes an object included in an image captured by a camera mounted on a vehicle. A first distance estimation unit estimates a distance between the vehicle and the recognized object based on the image. A second distance estimation unit estimates the distance between the vehicle and the recognized object based on the image by using an estimation method different from that of the first distance estimation unit. A combining unit combines a result of estimating the distance obtained by the first distance estimation unit and a result of estimating the distance obtained by the second distance estimation unit based on at least one of an amount of change in the distance estimated by the second distance estimation unit, a steering wheel angle of the vehicle, and information about an acceleration in an up-down direction, and outputs a result of the combination as a ranging result.