Stereo Processing Unit for 3D Vehicle Sensing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The high manufacturing cost and complexity of equipping self-driving vehicles with multiple sensing devices for 360-degree sensing, particularly the need for a quadrangle image sensor to project annular images, make it expensive to achieve three-dimensional sensing of the horizontal periphery.
Innovation Solution
A three-dimensional image processing device comprising two monocular cameras with fish-eye lenses, where one camera's imaging direction and region overlap with the other's to form common and non-common parts, allowing for stereo and motion stereo processing to generate range images, reducing the need for multiple stereo cameras and thus lowering costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple sensing devices including radar, laser scanner, and camera are provided for sensing in all directions, then sensing coverage is improved, but manufacturing cost and device complexity increase
Solution Approach 1:
The patent combines two monocular imaging units to perform stereo processing, merging their functions to achieve three-dimensional sensing. This reduces the need for multiple separate sensing devices while maintaining comprehensive sensing coverage in all directions around the vehicle.
Solution Approach 2:
The imaging units are configured to serve multiple functions: they capture images for stereo processing, perform motion stereo processing when the vehicle moves, and provide both two-dimensional and three-dimensional sensing capabilities. This multi-functionality reduces the overall number of devices needed.
2Adaptability or versatility
If a quadrangle image sensor is used to project annular image for sensing in all directions, then sensing coverage is improved, but manufacturing cost increases
Solution Approach 1:
The patent divides the sensing task into segments by using two monocular imaging units with fish-eye lenses that capture overlapping fields of view. Each unit covers a portion of the horizontal periphery, and their combined data provides complete 360-degree coverage without requiring a single complex quadrangle sensor.
Solution Approach 2:
The patent uses standard monocular cameras with fish-eye lenses, which are more readily available and less expensive than specialized quadrangle image sensors. These conventional imaging units achieve the required sensing functionality at lower manufacturing cost.
3Ease of manufacture
If two monocular imaging units with overlapping imaging regions are used for stereo processing, then manufacturing cost is reduced, but measurement precision may be compromised
Solution Approach 1:
The patent configures the imaging units so that their central imaging regions overlap to provide high-precision stereo processing for critical areas, while side regions provide additional coverage. This local optimization ensures measurement precision is maintained in the most important sensing zones.
Solution Approach 2:
The patent adjusts the baseline distance and imaging parameters of the two monocular units to optimize stereo processing accuracy. By carefully controlling the spatial relationship and imaging characteristics, measurement precision is maintained despite using cost-effective monocular cameras.
4Area of stationary object
If central parts of imaging regions are set as non-common parts without overlap, then field of view is expanded, but measurement precision in those regions is reduced
Solution Approach 1:
The patent employs motion stereo processing for the non-overlapping central regions, utilizing vehicle movement to synthesize depth information dynamically. This allows the system to maintain measurement precision in expanded field-of-view areas without requiring static overlapping coverage.
Data Source
AI summary
A three-dimensional image processing device includes: an input unit configured to acquire a first taken image and a second taken image respectively from a first imaging unit and a second imaging unit; and a stereo processing unit configured to execute stereo processing and then outputs a range image, for a common part where an imaging region of the first taken image and an imaging region of the second taken image have an overlap with each other, an imaging direction of the first imaging unit and an imaging direction of the second imaging unit are set toward a horizontal direction, and both side parts of the imaging region of the first imaging unit and both side parts of the imaging region of the second imaging unit are set as common parts.