Vehicular Vision Through Obstructed Body Areas Using AR View Synthesis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle imaging systems struggle to provide a clear view of obstructed areas, such as the vehicle's non-transparent portions, hindering the driver's visibility and maneuverability, especially during parking and low-speed maneuvers.
Innovation Solution
A vehicle vision system utilizing CMOS cameras and augmented reality through head-mounted displays like Google Glass, combined with eye and head tracking, allows the driver to virtually see through obstructed areas by projecting exterior images based on the driver's gaze direction, using virtual camera positions to create a transparent view.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If traditional vehicle imaging systems are used, then the system structure is simple, but the driver cannot see through non-transparent portions of the vehicle, reducing visibility and maneuverability
Solution Approach 1:
The system divides the vehicle's exterior view into multiple segments captured by separate cameras positioned at different locations (front, rear, sides). Each camera captures a specific field of view that would be obstructed by different portions of the vehicle body, and these segmented views are then synthesized to create a complete transparent view.
Solution Approach 2:
The system introduces an intermediary processing system that includes image processing units and a display device. This intermediary system captures raw images from multiple cameras, processes them to remove reflections and enhance clarity, and presents the processed information to the driver through a display, thereby mediating between the physical obstruction and the driver's perception.
2Loss of information
If multiple cameras are added to provide transparent view, then visibility of obstructed areas improves, but the system complexity and cost increase
Solution Approach 1:
The system employs universal image processing algorithms that can handle multiple camera inputs and various driving conditions. The same processing pipeline processes images from all camera positions, applying reflection removal, image stitching, and enhancement uniformly across all views, thereby reducing the need for position-specific processing hardware.
Solution Approach 2:
The system merges multiple camera feeds into a single integrated transparent view displayed on one display device. By combining the image data from all cameras and processing them together, the system achieves a comprehensive view while consolidating processing resources rather than requiring separate processing units for each camera.
3Measurement precision
If reflection removal processing is applied, then image quality of captured images improves, but processing time and computational load increase
Solution Approach 1:
The system applies reflection removal processing as a preliminary step in the image processing pipeline, before other operations like image stitching or enhancement. By removing reflections early in the processing sequence, subsequent processing steps work with cleaner base images, reducing the overall computational load and time required for complete image processing.
Data Source
AI summary
A vehicular vision system includes an interior-viewing camera disposed in a vehicle and viewing a driver of the vehicle, and a forward-viewing camera disposed at a windshield of the vehicle and viewing at least forward of the vehicle. Image data captured by the interior-viewing camera is transferred from the interior-viewing camera to and is processed at an electronic control unit (ECU). Image data captured by the forward-viewing camera is transferred to and is processed at the ECU. A hand position of the driver of the vehicle is determined via image processing at the ECU of image data captured by the interior-viewing camera. Image data captured by the forward-viewing camera is processed for at least one driver assistant system of the equipped vehicle. An object present exterior of the equipped vehicle is detected via image processing at the ECU of image data captured by the forward-viewing camera.