Monocular Navigation Assistance With Dual Depth of Field
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing navigation trajectory calculation methods for autonomous robots and vehicles face challenges in accurately estimating depth maps from monocular images, particularly in weakly textured scenes, and suffer from alignment issues when using multiple cameras or images with focus blur, which degrades other computer vision tasks.
Innovation Solution
A navigation assistance device and method using a single monocular camera that simultaneously acquires a sharp and a blurred image of a scene, leveraging a machine learning model to combine depth indices from both images for improved depth estimation without degrading other vision tasks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a camera focuses at a given plane to acquire an image with depth defocusing blur for improved depth estimation, then depth estimation performance is improved, but the image quality at all points degrades and cannot be exploited in other vision tasks
Solution Approach 1:
The patent divides the imaging function into two separate optical paths: one path (first lens) captures a sharp all-in-focus image for general vision tasks, while the other path (second lens) captures a blurred image with depth-dependent defocusing for depth estimation. This segmentation allows each optical path to be optimized for its specific function without compromising the other.
Solution Approach 2:
The patent creates a multi-functional imaging system where a single device simultaneously performs two functions: capturing sharp images for general computer vision tasks (segmentation, detection, localization) and capturing blurred images for depth estimation. The two optical paths work together to provide both functions from a single imaging system.
2Measurement precision
If two cameras are used to acquire one sharp image and one blurred image, then depth estimation can be improved, but alignment between the two images becomes difficult and introduces errors
Solution Approach 1:
The patent merges two optical paths into a single imaging device, where both the sharp and blurred images are captured simultaneously by different lenses (first lens and second lens) that are optically aligned to the same image sensor. This combining eliminates the alignment problems that would arise from using two separate cameras, as both images are inherently registered to the same coordinate system.
3Measurement precision
If several images focusing at different planes are used for depth estimation, then depth information can be obtained, but the acquisition requires alignment which is not possible with a single variable focal length camera on a mobile system
Solution Approach 1:
Instead of sequentially capturing images at different focal planes requiring alignment, the patent segments the imaging function into two simultaneous optical paths: one capturing a sharp all-in-focus image and the other capturing a blurred depth-encoded image. This eliminates the need for sequential acquisition and post-acquisition alignment.
4Device complexity
If monocular image approaches are used for depth estimation, then device complexity is reduced, but the transformation between image measurements and depth map becomes non-trivial and requires neural networks
Solution Approach 1:
The patent changes the optical parameter (depth of field) to encode depth information directly in the image formation process. By using a second lens with a smaller depth of field, the system creates depth-dependent defocusing blur that provides direct geometric cues about distance. This physical parameter change transforms the depth estimation problem from a complex computational task into a more straightforward process that leverages optical physics.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enhances depth estimation performance and reduces alignment errors, providing accurate navigation trajectories while maintaining image quality for other vision tasks like localization and segmentation.
Implementation Method 1
a monocular camera capable to simultaneously acquire the first image of the scene with a first depth of field and at least one second image of the scene with a second depth of field smaller than the first depth of field
Data Source
AI summary
A navigation assistance device intended to be embedded in a mobile system includes a monocular camera capable to simultaneously acquire a first image of a scene with a first depth of field and one or more second images of the scene with a second depth of field smaller than the first depth of field, a depth estimator that determines a depth map of the scene from the first image of the scene and the one or more second images of the scene, and a computer that calculates a navigation trajectory from the first image of the scene and the depth map of the scene.