Semantic Occupancy Grid Polygons for Faster ADAS Perception
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Advanced driver-assistance systems (ADAS) and autonomous driving systems face challenges in accurately processing environmental information to ensure vehicle safety, particularly in identifying objects within a safe braking distance and distinguishing between drivable and non-drivable areas, which affects the overall safety of occupants, pedestrians, and other road users.
Innovation Solution
The method involves generating an occupancy grid by receiving images, detecting objects, associating regions of interest, creating masks and boundary polygons, and reducing the number of vertices in these polygons to enhance processing efficiency and accuracy, thereby improving the system's ability to manage perception and navigation information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If detailed object detection and boundary polygon generation are performed for all detected objects, then measurement precision and reliability of occupancy grid are improved, but device complexity and processing power requirements increase
Solution Approach 1:
The occupancy grid is divided into multiple tiles, and processing is performed selectively on tiles containing detected objects rather than the entire grid. This segmentation allows high-precision object detection and boundary polygon generation to be applied only where necessary, reducing overall computational complexity while maintaining measurement precision for critical areas.
Solution Approach 2:
Different processing qualities are applied to different regions of the occupancy grid. High-precision object detection and detailed boundary polygons are generated for tiles containing detected objects, while lower-precision or simplified processing is applied to tiles without objects. This local quality approach maintains accuracy where needed while reducing device complexity in less critical regions.
2Manufacturing precision
If high-resolution masks and boundary polygons are generated for all regions, then manufacturing precision and measurement precision are improved, but loss of time and processing speed worsen
Solution Approach 1:
The processing area is segmented into tiles, and high-resolution mask and boundary polygon generation is performed only for tiles containing detected objects. This selective segmentation reduces the total number of polygons that need to be processed, thereby reducing processing time while maintaining boundary polygon accuracy for relevant regions.
Solution Approach 2:
Instead of generating high-resolution boundary polygons for the entire occupancy grid, the system applies partial action by generating detailed polygons only for regions containing detected objects. This partial processing approach maintains manufacturing precision for critical areas while significantly reducing the time loss associated with processing the entire grid at high resolution.
3Measurement precision
If the occupancy grid covers a large area with high resolution, then measurement precision and reliability are improved, but device complexity and processing power requirements increase
Solution Approach 1:
The large-area occupancy grid is segmented into multiple smaller tiles, allowing high-resolution processing to be applied to individual tiles rather than the entire grid at once. This segmentation reduces the memory and computational requirements for handling high-resolution data, thereby reducing device complexity while maintaining measurement precision through localized high-resolution processing.
Data Source
AI summary
An example driver assistance system includes an object detection (OD) network, a semantic segmentation network, a processor, and a memory. In an example method, an image is received and stored in the memory. An object detection (OD) polygon is generated for each object detected in the image, and each OD polygon encompasses at least a portion of the corresponding object detected in the image. A region of interest (ROI) is associated with each OD polygon. Such method may further comprise generating a mask for each ROI, each mask configured as a bitmap approximating a size of the corresponding ROI; generating at least one boundary polygon for each mask based on the corresponding mask, each boundary polygon having multiple vertices and enclosing the corresponding mask; and reducing a number of vertices of the boundary polygons based on a comparison between points of the boundary polygons and respective points on the bitmaps.