Stixel Detection Model for Multi-Resolution Camera Adaptability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing stixel and free space detection algorithms struggle to handle images with different resolutions, limiting their effectiveness in robotics and autonomous systems where cameras with varying resolutions are common.
Innovation Solution
A machine learning model, specifically a CNN-based approach, is adapted to process images in different resolutions without retraining, using multiresolution inference and ordinal regression to parameterize stixel and free space elements, allowing for robust detection across various camera resolutions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a stixel and free space detection algorithm is designed for a specific image resolution, then detection precision is improved for that resolution, but adaptability to different camera resolutions deteriorates
Solution Approach 1:
The patent applies parameter changes by transforming the detection problem into a parameter prediction task. Instead of directly detecting stixels at fixed resolution, the model predicts parameters (bottom position, height, depth) that can be applied to images of any resolution. This allows the same trained model to work across different camera resolutions without retraining, resolving the contradiction between detection precision and adaptability.
2Measurement precision
If separate models are created for each camera resolution, then detection precision for each resolution is improved, but device complexity and labeling efforts increase
Solution Approach 1:
The patent implements universality by designing a single detection model that can handle multiple camera resolutions simultaneously. The model predicts stixel parameters in a resolution-independent manner, allowing one model to serve multiple functions across different hardware configurations. This eliminates the need to create and maintain separate models for each resolution, reducing device complexity while maintaining detection precision.
3Measurement precision
If the model processes images at native resolution, then detection precision is improved, but computational resources and processing time increase
Solution Approach 1:
The patent extracts only the essential information needed for stixel detection by predicting specific parameters (bottom position, height, depth) rather than processing the entire image at full resolution. This parameter extraction approach maintains detection precision while significantly reducing computational resource requirements, as the model only needs to determine these key parameters rather than analyze every pixel.
Data Source
AI summary
The invention relates to a method (100) for generating an environment representation (90), comprising the following steps:providing (101) at least one image (40) which results from a recording by at least one image-capturing device (25) and shows at least one object (60) and/or a navigable region (65) in an environment of the at least one image-capturing device (25), the image (40) being split into a plurality of image columns (41),generating (102) the environment representation (90), one stixel and/or free space element (30) of the respective image column (41) of the at least one provided image (40) being parameterized for this purpose in order to represent the object (60) and/or the navigable region (65).