Virtual Rounded Cuboid Distance Estimation for Autonomous Vehicles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for a moving body, such as autonomous vehicles, face challenges in accurately estimating distance to objects due to limitations in resolving distance information within the inner region of virtual spheres with smallest radii, particularly when the moving body's horizontal and vertical axes have different lengths.
Innovation Solution
A method utilizing multiple cameras with wide Field of Views, arranged to cover all directions, projects images onto virtual rounded cuboids with extended planes and curved surfaces, applying 3D concatenation and cost volume computation networks to generate inverse radius indices, enabling accurate distance calculation from the moving body to objects in any direction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If virtual spheres with different radii are used to obtain distance information, then distance measurement coverage is improved, but measurement precision deteriorates for objects within the inner region of the smallest virtual sphere
Solution Approach 1:
The patent replaces virtual spheres with virtual rounded cuboids that have curved surfaces. This curvature change allows the geometric model to better fit and represent the inner region space, enabling accurate distance measurement for objects located within the inner region while maintaining coverage for objects at various distances.
Solution Approach 2:
The patent transitions from spherical geometry to rounded cuboid geometry, adding angular dimensionality to the distance representation. This dimensional change enables the system to distinguish between different angular positions and distances more precisely, resolving the ambiguity in the inner region.
2Measurement precision
If LIDAR or RADAR sensors are used to obtain distance information, then distance measurement capability is improved, but device complexity and power consumption increase
Solution Approach 1:
The patent creates virtual rounded cuboid models that replicate the functional capability of LIDAR/RADAR distance measurement using only camera images. This virtual modeling approach eliminates the need for complex active sensors while achieving similar distance estimation capabilities through image processing.
Solution Approach 2:
The patent replaces active mechanical sensing systems (LIDAR/RADAR) with passive optical sensing (cameras). By substituting the mechanical/optical active systems with a computational approach using standard cameras, the system reduces device complexity and power consumption while maintaining distance measurement functionality.
3Device complexity
If conventional camera-based virtual sphere projection is used, then device simplicity is maintained, but adaptability to objects in inner region deteriorates
Solution Approach 1:
The patent modifies the virtual projection geometry from spheres to rounded cuboids with curved surfaces. This geometric modification enables the system to adapt to and accurately represent objects in the inner region while maintaining the simplicity of the camera-based approach.
Solution Approach 2:
The patent changes the geometric parameters of the virtual projection model from spherical to rounded cuboidal. This parameter change transforms the system's ability to handle inner region objects by adjusting the projection geometry without adding complex hardware.
Data Source
AI summary
A method for acquiring a distance from a moving body to an object located in any direction of the moving body includes steps of: an image processing device (a) instructing a rounded cuboid sweep network to project pixels of images, generated by cameras covering all directions of the moving body, onto N virtual rounded cuboids to generate rounded cuboid images and apply 3D concatenation operation thereon to generate an initial 4D cost volume, (b) instructing a cost volume computation network to generate a final 3D cost volume from the initial 4D cost volume, and (c) generating inverse radius indices, corresponding to inverse radii representing inverse values of separation distances of the N virtual rounded cuboids, by referring to the final 3D cost volume and extracting the inverse radii by using the inverse radius indices, to acquire the separation distances and thus, the distance from the moving body to the object.