Object Bounding Contours for Accurate Autonomous Vehicle Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing bounding boxes in autonomous vehicles fail to accurately represent the size and shape of objects due to their rectangular nature, often expanding to include protrusions, leading to inaccurate representations and potential safety hazards.
Innovation Solution
Generating bounding contours as geometric shapes, such as polygons or freeform shapes, that closely follow the outer edges of objects, including irregularities, using image and lidar data to provide a more accurate representation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If rectangular bounding boxes are used to represent objects, then the representation is simple and easy to process, but the accuracy of object size and shape representation deteriorates
Solution Approach 1:
The patent segments the object boundary into multiple discrete points along its perimeter. Instead of using a single rectangular bounding box, the system divides the object contour into numerous small segments (boundary points) that collectively form a precise representation of the object's irregular shape. This segmentation allows the system to capture detailed geometric information while maintaining computational tractability through discrete point processing.
Solution Approach 2:
The patent transitions from two-dimensional rectangular bounding boxes to multi-dimensional polygonal representations by incorporating multiple boundary points with coordinates (x, y, z) in three-dimensional space. This dimensional expansion enables the representation of complex, irregular object shapes that cannot be accurately captured by simple 2D rectangles, while still allowing for efficient computational processing through coordinate-based mathematics.
2Reliability
If rectangular bounding boxes are expanded to include protrusions, then all parts of the object are covered, but the accuracy of object representation deteriorates due to including extra space
Solution Approach 1:
The patent applies local quality by allowing different regions of the object boundary to be represented with different levels of detail. Instead of forcing a uniform rectangular approximation, the system places boundary points specifically at locations where the object's shape deviates from rectangular form (such as protrusions and indentations). This localized refinement ensures that each part of the object is represented with appropriate precision, capturing local geometric features without unnecessarily complicating the overall representation.
Solution Approach 2:
The patent introduces dynamic adaptability by allowing the bounding representation to adjust its complexity based on the object's geometry. The system dynamically determines the number and placement of boundary points along the object contour, increasing density in regions with complex geometry (protrusions, irregular shapes) and using fewer points in regions with simple, regular boundaries. This dynamic approach optimizes the balance between coverage completeness and representation accuracy.
Data Source
AI summary
Techniques are discussed herein for controlling autonomous vehicles within a driving environment, including generating and using bounding contours associated with objects detected in the environment. Image data may be captured and analyzed to identify and/or classify objects within the environment. Image-based and/or lidar-based techniques may be used to determine depth data associated with the objects, and a bounding contour may be determined based on the object boundaries and associated depth data. An autonomous vehicle may use the bounding contours of objects within the environment to classify the objects, predict the positions, poses, and trajectories of the objects, and determine trajectories and perform other vehicle control actions while safely navigating the environment.