Slope Estimation via Image Segmentation and Pitch Angle Calculation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional distance recognition systems in self-driving vehicles struggle to accurately estimate the slope of roads, especially on uphill or downhill roads, and fail to account for curved roads where the vanishing point is undefined, making it difficult to determine the pitch angle of the vehicle.
Innovation Solution
An apparatus using at least one camera and an image signal processor that divides the forward image into smaller regions, identifies lane segments, obtains coordinate values, and calculates the pitch angle of each region, allowing for the conversion of image coordinates to world coordinates, which is then used to control the vehicle's suspension and speed based on the road's slope.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional distance recognition assumes vehicles are on the same plane, then the system is simple to implement, but it fails to accurately estimate slope on uphill or downhill roads
Solution Approach 1:
The forward image is divided into multiple smaller regions in the horizontal direction, with each region processed independently to obtain local pitch angles. This segmentation allows the system to handle complex road geometries by breaking them down into manageable segments, improving slope estimation accuracy without overwhelming computational resources.
Solution Approach 2:
The patent transforms the problem from 2D image coordinates to 3D world coordinates by calculating pitch angles and using them to convert image coordinates into world coordinates. This dimensional transformation enables accurate slope estimation on curved and sloped roads by incorporating the pitch angle as an additional parameter in the coordinate conversion process.
2Adaptability or versatility
If the system uses vanishing point for pitch angle estimation, then the method is simple, but it cannot work on curved roads where vanishing point is undefined
Solution Approach 1:
By dividing the image into multiple smaller regions and processing each region independently, the system can handle curved roads where the overall vanishing point is undefined. Each local region may still have a definable vanishing point or lane geometry that can be used to estimate the pitch angle for that specific segment, thereby maintaining adaptability across different road types.
Solution Approach 2:
The patent applies different processing approaches to different regions of the image based on local road characteristics. For each smaller region, the system identifies lane segments and calculates pitch angles based on local geometry, allowing the system to adapt to varying road conditions (straight, curved, sloped) in different parts of the field of view while maintaining overall system accuracy.
3Measurement precision
If the system divides the image into multiple smaller regions and processes each independently, then slope estimation accuracy improves, but computational load increases
Solution Approach 1:
The image is divided into multiple smaller regions that are processed in parallel, which improves measurement accuracy by capturing local road geometry details. The segmentation strategy balances computational load by creating regions of manageable size that can be processed efficiently while still providing sufficient resolution for accurate pitch angle estimation.
Solution Approach 2:
The system processes multiple smaller regions rather than the entire image at once, applying partial action to each region. This approach achieves accurate slope estimation by focusing computational resources on relevant local features (lane segments in each region) rather than processing the entire image uniformly, thereby reducing overall computational energy consumption while maintaining precision.
Data Source
AI summary
An operating method of a slope estimating apparatus is provided. The operating method of the slope estimating apparatus including at least one camera includes obtaining a forward image through the at least one camera, detecting a lane included in the forward image, dividing the forward image into a plurality of smaller regions in a horizontal direction, identifying a plurality of lane segments included in each of the plurality of smaller regions, obtaining a plurality of coordinate values forming each of the plurality of lane segments, and obtaining a pitch angle of each of the plurality of smaller regions based on the obtained plurality of coordinate values.