Virtual Vehicle Transparency for Surround View Obstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Surround view camera systems in vehicles often obscure pedestrians and objects behind a virtual representation of the vehicle, making them difficult to detect and track in the simulated image display.
Innovation Solution
Implementing object detection methods like neural network computer vision algorithms and sensor fusion to classify objects and predict their motion, then modifying the virtual vehicle's transparency based on obstruction levels and visibility, with features like highlight frames and depth of field effects to enhance object recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a virtual image of the vehicle is added to the synthesized view to provide realism, then the visual representation becomes more realistic and complete, but objects behind the virtual vehicle become obscured and difficult to detect
Solution Approach 1:
The virtual vehicle image is rendered with spatially varying transparency properties. Specifically, regions of the virtual vehicle that would obscure important objects (detected through object detection algorithms) are made transparent or semi-transparent, while other regions maintain full opacity to preserve realism. This local differentiation of transparency quality allows the system to maintain overall realism while selectively revealing obscured objects.
Solution Approach 2:
The system modifies the visual properties of the virtual vehicle image by changing its transparency (optical property) in response to detected objects. When an object is detected behind a portion of the virtual vehicle, that portion's transparency is adjusted to allow the obscured object to be visible. This dynamic property modification resolves the contradiction between maintaining realism (through the virtual vehicle image) and enabling object detection.
2Difficulty of detecting and measuring
If the virtual vehicle image is made transparent to reveal obscured objects, then object detectability improves, but the realism and visual completeness of the virtual image deteriorates
Solution Approach 1:
Rather than making the entire virtual vehicle image transparent (which would destroy realism), the system applies transparency selectively to only those specific regions where objects are obscured. This localized approach maintains the overall realism and visual completeness of the virtual vehicle while enabling detection of critical obscured objects.
Solution Approach 2:
The system applies transparency partially - only to the extent necessary to reveal obscured objects, rather than applying it uniformly across the entire virtual vehicle image. This partial application of transparency achieves the goal of object detection while minimizing the loss of realism.
3Difficulty of detecting and measuring
If object detection algorithms are used to identify obscured objects, then the ability to track objects improves, but the system complexity increases
Solution Approach 1:
The system uses computer vision algorithms and object detection techniques as intermediary processing steps between the raw camera images and the final rendered view. These algorithms analyze the images to detect objects, determine their locations, and identify which objects are obscured by the virtual vehicle. This intermediary detection layer enables intelligent transparency adjustments without requiring complex hardware modifications.
Solution Approach 2:
The system replaces potential mechanical or hardware-based solutions (such as multiple physical cameras or complex optical systems) with software-based computer vision and image processing algorithms. By using neural network algorithms and image analysis techniques, the system achieves sophisticated object detection and tracking capabilities through computational methods rather than mechanical complexity.
Data Source
AI summary
A visual scene around a vehicle is displayed to an occupant of the vehicle on a display panel as a virtual three-dimensional image from an adjustable point of view outside the vehicle. A simulated image is assembled corresponding to a selected vantage point on an imaginary parabolic surface outside the vehicle from exterior image data and a virtual vehicle image superimposed on a part of the image data. Objects are detected at respective locations around the vehicle subject to potential impact. An obstruction ratio is quantified for a detected object having corresponding image data in the simulated image obscured by the vehicle image. When the detected object has an obstruction ratio above an obstruction threshold, a corresponding bounding zone of the vehicle image is rendered at least partially transparent in the simulated image to unobscure the corresponding image data.