Object Detection Using Orientation Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting target objects in images, particularly in bin picking scenarios, face challenges such as incorrect detection due to occlusion and lighting issues, and prioritize objects based on partial area features rather than overall visibility, leading to high calculation costs and unreliable results.
Innovation Solution
An information processing apparatus that estimates and identifies the orientation of detection candidate objects in measurement areas, determines the best detection target based on correspondence between estimated and identified orientations, using Randomized Tree classification and voting processes to prioritize objects with larger visible areas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If detection is performed using entire contour information and partial contour information multiple times, then occlusion detection accuracy is improved, but calculation cost increases
Solution Approach 1:
The patent divides the object detection process into two stages: first detecting the entire object contour, then dividing the object into multiple partial areas for separate detection. This segmentation allows the system to evaluate occlusion in different regions systematically without requiring excessive computational resources for the entire object at once.
Solution Approach 2:
The patent applies different detection strategies to different partial areas of the object. By setting multiple measurement areas across the object and evaluating detection likelihood separately for each area, the system can identify regions with strong features versus weak features, and handle occlusion detection with appropriate precision for each local region rather than applying uniform high-cost processing everywhere.
2Productivity
If priority is determined based only on occlusion in set partial areas, then detection speed is improved, but detection reliability deteriorates
Solution Approach 1:
The patent extends the evaluation from two-dimensional partial area detection to three-dimensional orientation verification. By estimating the orientation of the object and checking whether detected partial areas correspond to the expected three-dimensional structure, the system adds a dimensional check that improves reliability without significantly slowing down the detection process.
Solution Approach 2:
The patent implements a feedback mechanism where the detection results from multiple partial areas are combined to verify the overall object detection. The system checks whether the detected partial areas are consistent with each other and with the estimated object orientation, providing a feedback loop that ensures detection reliability while maintaining efficient processing.
Data Source
AI summary
When there are a plurality of detection candidate objects, to detect a target object that is appropriately visible as a whole, an information processing apparatus calculates detection likelihoods in a plurality of local areas of each of the plurality of detection target candidates, and a detection reliability of each of the detection target candidates based on a distribution of the detection likelihoods.