Multi-resolution image patches for autonomous navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional autonomous vehicle systems face challenges in accurately predicting driving paths due to high computational loads from processing high-resolution images, while low-resolution images may insufficiently capture road features at a distance, impairing vehicle safety.
Innovation Solution
The use of multi-resolution image patches, where portions of the image are captured at different resolutions, allows a Deep Neural Network (DNN) to perceive the environment clearly at a distance while maintaining sufficient perception up close, reducing computational load and improving path prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution image data is used by the DNN, then the perception accuracy of the environment at a distance is improved, but the computational load increases
Solution Approach 1:
The image data is divided into multiple patches, where at least one patch is a high-resolution patch and others are low-resolution patches. This segmentation allows the DNN to process only specific regions at high resolution, improving perception accuracy for critical areas while reducing overall computational load by processing other areas at lower resolution.
Solution Approach 2:
Different resolution qualities are applied to different regions of the image based on their importance. High-resolution patches are allocated to regions requiring detailed perception (such as areas with potential obstacles or road features), while low-resolution patches are used for less critical regions. This local quality differentiation optimizes both perception accuracy and computational efficiency.
2Power
If low-resolution image data is used by the DNN, then the computational load is reduced, but the perception accuracy of road features at a distance deteriorates
Solution Approach 1:
The image is segmented into multiple patches with different resolution levels. By identifying and designating specific patches as high-resolution, the system ensures that critical road features and obstacles at a distance are captured with sufficient detail, while non-critical areas use lower resolution to reduce computational load.
Solution Approach 2:
The system applies varying quality levels (resolutions) to different local regions of the image based on their perceptual importance. Critical regions containing road features, obstacles, or navigation-relevant information are processed at high resolution, while other regions use lower resolution, thus balancing computational efficiency with necessary perception accuracy.
3Measurement precision
If uniform high-resolution image data is used, then the DNN can accurately perceive distant objects, but the processing time and computational resources increase
Solution Approach 1:
The image is divided into multiple patches that are processed independently or selectively by the DNN. At least one high-resolution patch ensures accurate perception of distant objects, while other low-resolution patches reduce the total number of pixels requiring processing, thereby decreasing processing time and computational resource consumption.
Solution Approach 2:
Different quality levels are applied locally across the image based on the importance of each region for navigation and safety. High-resolution processing is concentrated on regions containing distant objects of interest, while other regions use lower resolution, optimizing the balance between perception accuracy and processing efficiency.
4Productivity
If uniform low-resolution image data is used, then the processing efficiency is improved, but the ability to extract road features at far distance deteriorates
Solution Approach 1:
The image is segmented into multiple patches with differentiated resolution levels. High-resolution patches are strategically assigned to regions containing critical road features and obstacles at a distance, ensuring accurate feature extraction, while low-resolution patches maintain processing efficiency for less critical areas.
Solution Approach 2:
The system applies local quality variation by processing different image regions at different resolutions. Regions containing road features, lane markings, obstacles, or other navigation-critical elements are processed at high resolution to ensure accurate extraction, while other regions use lower resolution to maintain overall processing efficiency.
Data Source
AI summary
In examples, image data representative of an image of a field of view of at least one sensor may be received. Source areas may be defined that correspond to a region of the image. Areas and/or dimensions of at least some of the source areas may decrease along at least one direction relative to a perspective of the at least one sensor. A downsampled version of the region (e.g., a downsampled image or feature map of a neural network) may be generated from the source areas based at least in part on mapping the source areas to cells of the downsampled version of the region. Resolutions of the region that are captured by the cells may correspond to the areas of the source areas, such that certain portions of the region (e.g., portions at a far distance from the sensor) retain higher resolution than others.