Robot Object Recognition Feedback for Misidentification Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Robots often misrecognize objects, leading to unnecessary operations and reduced utility in interactions with users, such as approaching unwanted objects or misidentifying animals as humans.
Innovation Solution
A robot equipped with an image sensor, driving part, memory, and processor that uses object recognition models to compare images captured at different distances and angles to determine object misrecognition, storing error information and adjusting operation controls based on reliability and similarity analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the robot uses object recognition model to identify objects, then the robot can interact with users, but the robot may misrecognize objects leading to unnecessary operations
Solution Approach 1:
The robot captures a first image before approaching the object and performs preliminary object recognition. Based on the recognition result and reliability assessment, the robot decides whether to proceed with approaching the object, thereby preventing unnecessary operations before they occur.
Solution Approach 2:
The robot captures a second image while approaching the object and compares it with the first image. By evaluating changes in recognition results and reliability values between the two images, the robot feedback-adjusts its operation decisions, confirming or canceling the approach action based on consistent recognition across different viewing conditions.
2Reliability
If the robot captures multiple images and compares recognition results, then object misrecognition is reduced, but the operation time increases
Solution Approach 1:
The robot performs object recognition partially - capturing a first image for preliminary identification and only conditionally capturing a second image when the first recognition meets reliability criteria. This partial action approach achieves sufficient recognition accuracy without always performing the complete two-image verification sequence.
Solution Approach 2:
The robot autonomously determines whether to perform the second image capture and comparison based on the reliability of the first recognition result. When the first recognition is sufficiently reliable, the robot self-services by proceeding directly without the additional verification step, thereby saving time while maintaining accuracy.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A robot is provided. The robot includes an image sensor, a driving part, a memory configured to store at least one instruction, and a processor, and the processor is configured to obtain a first image captured by the image sensor, obtain first information on an object included in the first image, control the driving part for the robot to approach the object, obtain a second image while the robot is approaching the object, obtain second information on an object based on the second image, determine whether an error is present in first information or second information based on the first information and the second information, and control an operation of the robot based on a determination result.