Object Detection Region Prioritization and Re-capture
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection systems in mixed or augmented reality face challenges in achieving high accuracy without significant increases in model complexity, resource consumption, or overcoming photographic errors.
Innovation Solution
An object detection device and method that utilize a controller to receive classification data and image regions, prioritize regions, determine if re-capture is needed based on photographic features, and update the image capturing device to re-process and re-generate classification data, thereby improving detection accuracy with minimal computational overhead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning models with larger model size are used to improve object detection accuracy, then detection accuracy is improved, but resource consumption (power, memory, processing) increases significantly
Solution Approach 1:
The patent divides the image into multiple regions and processes each region separately through the object detection model. This segmentation allows the system to reduce the overall computational load by processing smaller regional images rather than the entire high-resolution image, thereby maintaining detection accuracy while reducing resource consumption.
Solution Approach 2:
The patent applies different processing qualities to different regions of the image. Regions containing objects of interest are processed with higher quality and detail, while other regions use lower processing quality. This local quality approach maintains detection accuracy for critical areas while reducing overall computational resources required.
2Use of energy by moving object
If classic machine learning approaches with feature definition are used, then resource consumption is reduced, but object detection accuracy falls below human accuracy even for simple tasks
Solution Approach 1:
The patent performs preliminary processing of the image by dividing it into regions and identifying areas of interest before applying the object detection model. This preliminary action prepares the data in a way that enhances the effectiveness of subsequent detection processes, enabling accurate detection with reduced computational resources.
Solution Approach 2:
The patent replaces traditional mechanical feature engineering approaches with a learned feature extraction system that automatically identifies relevant features from the divided image regions. This substitution enables the system to achieve deep learning-level accuracy while operating on segmented regions that reduce overall computational burden.
3Productivity
If a two-step approach with coarse grain analysis is used to save processing power, then processing efficiency is improved, but accuracy deteriorates when objects are difficult to detect
Solution Approach 1:
The patent implements a dynamic processing approach where the level of detail and processing intensity is adjusted based on the content of each region. Regions containing difficult-to-detect objects receive higher processing intensity and detail, while simpler regions use lower processing intensity. This dynamic adaptation maintains accuracy for challenging objects while preserving overall processing efficiency.
Data Source
AI summary
An object detection device has a controller configured to receive classification data and regions for an image. The regions are prioritized and if a region is selected for further prediction, a photographic feature for the selected region is determined and a re-capture of the selected region is made based on the photographic feature by updating an image capturing device. The re-captured region is re-processed and classification data is re-generated. Then, final object detection for selected region(s) is performed and final object detection for not-selected region(s) is performed.