Endoscope Viewpoint Estimation With Unknown-Region Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing endoscopy systems face reduced accuracy in estimating the position and posture of an endoscope when an unknown structure region is present in the endoscopic image, leading to a decrease in the amount of information obtained and impaired estimation.
Innovation Solution
An information processing apparatus that detects unknown structure regions in endoscopic images, acquires a three-dimensional model of the body, estimates the endoscope's viewpoint using machine learning models, and generates virtual endoscopic images to enhance estimation accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If unknown structure regions are present in endoscopic images, then the endoscope can capture a broader field of view, but the accuracy of position and posture estimation is reduced
Solution Approach 1:
The patent extracts and removes unknown structure regions from the endoscopic image before processing. The detection unit identifies regions with low confidence in object recognition, and the estimation unit excludes these regions from the feature point matching process, thereby maintaining accurate estimation despite the presence of unknown structures in the original image.
Solution Approach 2:
The patent applies different processing quality to different regions of the image. Known structure regions are processed with high precision for feature extraction, while unknown structure regions are either excluded or processed with lower weight. This local differentiation allows the system to maintain overall estimation accuracy without being compromised by uncertain regions.
2Measurement precision
If unknown structure regions are excluded from processing, then estimation accuracy is maintained, but the amount of information utilized is reduced
Solution Approach 1:
The patent introduces an intermediary processing stage that separates known and unknown structure regions. The detection unit acts as an intermediary to identify and classify regions, allowing the estimation unit to process only reliable information while still accounting for the spatial context provided by unknown regions through the three-dimensional model.
Solution Approach 2:
The patent segments the endoscopic image into known structure regions and unknown structure regions. By dividing the image data into distinct segments with different processing priorities, the system can utilize reliable information from known regions for accurate estimation while disregarding or lightly processing unknown regions, thus balancing accuracy with information utilization.
3Device complexity
If feature point matching is performed using all detected features, then processing is simplified, but estimation accuracy decreases due to unreliable features from unknown regions
Solution Approach 1:
The patent performs preliminary action by detecting and classifying structure regions before the feature point matching process. The detection unit pre-identifies unknown structure regions and marks them for exclusion, so that the subsequent estimation unit only processes reliable features from known regions, simplifying the matching process while maintaining accuracy.
Solution Approach 2:
The patent extracts and removes unreliable feature points from unknown structure regions before performing feature point matching. This extraction step ensures that only features from known structure regions are used for estimation, simplifying the matching process to focus on reliable data while maintaining high estimation accuracy.
Data Source
AI summary
An information processing apparatus detects an unknown structure region from an endoscopic image, acquires a bronchial model generated in advance, estimates a viewpoint of an endoscope at a position where the endoscopic image has been captured, using the endoscopic image, the unknown structure region, and the bronchial model, and notifies of the estimated viewpoint of the endoscope.