Multi-Cuboid Heading Estimation for Under-Segmented Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous vehicle systems face challenges in accurately estimating cuboid geometries and headings due to errors in classifier detection, particularly from under-segmentation issues.
Innovation Solution
The system generates cuboids using a combination of image-based, lidar-based, and heuristic methods, employing a machine learning model to estimate object headings and generate bounding box geometries, which are used for robotic system movements such as object and vehicle trajectory planning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single cuboid estimation method (image-based, lidar-based, or heuristic) is used, then the system complexity is low, but the accuracy of cuboid estimation and heading determination deteriorates due to classifier detection errors and under-segmentation issues
Solution Approach 1:
The patent combines multiple cuboid estimation methods (image-based, lidar-based, and heuristic approaches) into a unified system. The machine learning model integrates outputs from these different techniques to generate a final cuboid estimation, thereby improving accuracy while managing system complexity through structured integration
2Measurement precision
If multiple cuboid estimation methods are combined, then the accuracy of heading determination is improved, but the computational complexity and processing time increase
Solution Approach 1:
The system performs preliminary cuboid estimations using multiple methods (image-based, lidar-based, heuristic) before the final machine learning model processing. This preliminary action allows the ML model to work with pre-processed data from multiple sources, improving heading accuracy while optimizing computational efficiency
3Ease of operation
If classifier detection is used for object identification, then the system operation is simplified, but under-segmentation errors occur leading to inaccurate cuboid geometries
Solution Approach 1:
The patent introduces an intermediary machine learning model that processes outputs from multiple cuboid estimation methods. This intermediary component reconciles the simplified classifier detection approach with the need for accurate cuboid geometries, reducing under-segmentation errors while maintaining operational simplicity
Data Source
AI summary
Disclosed herein are systems, methods, and computer program products for operating a robotic system. For example, the method includes: obtaining a first cuboid generated based on an image, a second cuboid generated based on a lidar dataset and/or a third cuboid generated by a heuristic algorithm using the lidar dataset; using a machine learning model to generate a heading for an object in proximity to the robotic system based on the first cuboid, second cuboid and/or third cuboid; generating a bounding box geometry and a bounding box location based on the second cuboid or third cuboid; and generating a fourth cuboid using the bounding box geometry, the bounding box location, and the heading generated using the machine learning model.