Robot Object Detection During Head Rotation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection systems in robots, such as humanoid communication robots, face challenges in efficiently detecting objects, particularly faces, when the robot's head is rotating, as conventional methods require completion of rotation for accurate face detection, leading to delayed response times.
Innovation Solution
The implementation of a distributed face detection process that determines an overlapping area between images captured at different times during rotation and performs face detection on areas outside this overlap, allowing for real-time object detection and improved responsiveness by initiating face detection before the head's rotation is complete.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If face detection is performed only after head rotation is complete, then detection accuracy is improved, but response time increases
Solution Approach 1:
The system performs preliminary face detection on the overlapping area during head rotation, before the imaging unit completes its rotation. This preliminary detection identifies faces that are likely to be in the final field of view, allowing the system to prepare detection results in advance and reduce overall response time while maintaining accuracy through verification after rotation completes.
Solution Approach 2:
The detection process is segmented into two phases: preliminary detection during rotation on the overlapping area, and final verification after rotation completes. This segmentation allows parallel processing of detection tasks, where the preliminary detection runs concurrently with head rotation, effectively utilizing time that would otherwise be wasted and reducing total response time.
2Loss of time
If distributed face detection is performed during head rotation, then response time is reduced, but processing complexity increases
Solution Approach 1:
The detection process is divided into preliminary detection during rotation and final verification after rotation. This segmentation manages complexity by handling different detection tasks at appropriate times, reducing the peak processing load while maintaining accuracy through the two-phase approach.
Solution Approach 2:
Preliminary detection is performed on the overlapping area during head rotation, preparing detection results in advance. This preliminary action reduces the amount of work needed after rotation completes, effectively distributing processing load over time and reducing peak complexity while improving response time.
3Measurement precision
If the robot waits for head rotation to complete before detecting objects, then detection accuracy is improved, but productivity decreases
Solution Approach 1:
The system performs preliminary object detection during head rotation on areas that are likely to contain objects of interest. This preliminary action enables the robot to prepare detection results and begin processing before the imaging unit completes its rotation, improving interaction responsiveness while maintaining detection accuracy through subsequent verification.
Solution Approach 2:
The detection process continues throughout the head rotation rather than waiting for it to complete. By performing preliminary detection during rotation and continuing verification after rotation, the system maintains continuous useful action, maximizing productivity and responsiveness without sacrificing detection accuracy.
Data Source
AI summary
The present invention is to reduce the time required to detect an object after completion of the rotation of a head or a body of a robot. A robot 100 includes a camera 111, and a control unit 127 which determines an overlapping area between a first image captured with the camera 111 at first timing and a second image captured with the camera 111 at second timing later than the first timing to detect an object included in an area of the second image other than the determined overlapping area.