Pose Tracking Bypasses Item Recognition Load
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Mobile computing devices face challenges in efficiently processing images of items in facilities due to the complexity of item recognition, leading to incomplete or missed item recognitions, especially when dealing with multiple items in a single image.
Innovation Solution
The device employs pose tracking and historical data to bypass computationally intensive item recognition mechanisms by correlating regions of interest in an image with previously recognized item identifiers, allowing it to derive or retrieve item identifiers efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a recognition mechanism is executed to derive item identifiers from all regions of interest in an image, then item recognition accuracy is improved, but computational load increases and processing speed decreases
Solution Approach 1:
The system performs preliminary actions by tracking the device pose and predicting item locations before executing the recognition mechanism. By using the tracked pose to determine where items are likely to be located, the system prepares in advance which regions need recognition processing, avoiding the need to process all regions in the image and thus reducing computational load while maintaining accuracy.
Solution Approach 2:
The system applies local quality by focusing computational resources only on specific regions of interest rather than processing the entire image uniformly. By using pose tracking to identify which local regions contain items, the recognition mechanism is applied selectively to those specific areas, improving processing efficiency while maintaining recognition accuracy for actual items.
2Reliability
If the recognition mechanism is executed for every region of interest, then item detection completeness is improved, but computational resources are excessively consumed
Solution Approach 1:
The system performs preliminary pose tracking and item location prediction before recognition processing. This preliminary action identifies which regions actually contain items based on the tracked device pose, allowing the system to skip recognition processing for regions where no items are expected, thus reducing computational resource consumption while maintaining detection completeness for actual items.
Solution Approach 2:
The system applies partial action by executing the recognition mechanism only for a subset of regions of interest that are predicted to contain items, rather than processing all regions. This selective approach reduces computational resource consumption while maintaining detection completeness by focusing resources on the most relevant areas identified through pose tracking and prediction.
3Productivity
If pose tracking and historical data are used to bypass recognition mechanisms, then computational load is reduced and processing speed is improved, but item recognition accuracy may deteriorate
Solution Approach 1:
The system performs preliminary pose tracking and item location prediction to identify which regions contain items before recognition processing. This preliminary action provides accurate guidance for where recognition should be applied, ensuring that even when bypassing recognition for some regions, the system maintains accuracy by focusing recognition resources on the correct locations based on tracked pose information.
Solution Approach 2:
The system uses feedback from pose tracking and historical data to continuously refine item location predictions. By incorporating feedback from the tracked device pose and previously identified item locations, the system improves the accuracy of predicting where items are located, ensuring that recognition processing is applied to the correct regions and maintaining overall recognition accuracy while reducing computational load.
Data Source
AI summary
A method in a mobile computing device includes: controlling a camera to capture an image; tracking, in association with the image, a pose of the mobile computing device in a coordinate system; detecting a region of interest (ROI) depicting an item in the image; determining a location of the ROI in the coordinate system, based on the tracked pose; obtaining an item identifier corresponding to the ROI by (i) when a previously recognized item identifier is not available, executing a recognition mechanism to derive the item identifier from the ROI, and (ii) when a previously recognized item identifier is available, bypassing the recognition mechanism and retrieving the previously recognized item identifier; and returning the obtained item identifier corresponding to the ROI.