Occlusion Detection via Keypoint Depth and Confidence Changes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing occlusion detection methods for medical imaging rely on the shape, color, pattern, or training data specific to occlusions, limiting their applicability and requiring extensive data collection and re-labeling for new occlusions.
Innovation Solution
The method involves acquiring an image sequence of an object and determining the occlusion state of keypoints based on confidence level change information and depth change information, independent of occlusion shape, color, or pattern.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If existing occlusion detection methods rely on shape, color, pattern, or training data specific to occlusions, then detection can be performed for known occlusion types, but applicability is limited and extensive data collection and re-labeling is required for new occlusions
Solution Approach 1:
The patent changes the detection parameters from occlusion-specific features (shape, color, pattern) to motion-based parameters (depth change information, confidence level change information). This allows the system to detect any occlusion regardless of its visual characteristics, eliminating the need for re-collecting and re-labeling training data when encountering new occlusion types.
Solution Approach 2:
The patent creates a universal occlusion detection method that works across all occlusion types by using motion-based detection. The same detection algorithm can identify coils, blankets, complex clothing, masks, or any other occlusion without requiring specific training data for each type, achieving multi-functionality in occlusion detection.
2Reliability
If occlusion detection is performed using traditional methods, then detection can be achieved for specific occlusion types, but accuracy and reliability are reduced when encountering unseen occlusion patterns
Solution Approach 1:
The patent performs preliminary action by acquiring an image sequence over time and establishing baseline confidence levels and depth information before detecting occlusions. This allows the system to compare current state with historical data, improving reliability by detecting occlusions based on deviations from normal patterns rather than relying on pre-trained recognition of specific occlusion types.
Solution Approach 2:
The patent uses feedback mechanisms by continuously monitoring confidence level changes and depth changes across the image sequence. The system feeds back the detected occlusion state to adjust positioning information, creating a closed-loop detection system that improves reliability through continuous verification and adaptation to actual conditions.
3Measurement precision
If confidence level change information and depth change information are used to determine occlusion state, then accuracy and reliability of occlusion detection are improved, but computational complexity increases
Solution Approach 1:
The patent segments the detection process into distinct computational stages: acquiring image sequence, extracting depth information, determining confidence levels, calculating changes, and comparing against thresholds. This segmentation makes the complex algorithm more manageable and implementable by breaking it into modular processing steps that can be executed efficiently.
Data Source
AI summary
Provided in embodiments of the present application are an occlusion detection method for medical imaging, a medical imaging method, and a medical imaging system. The occlusion detection method for medical imaging includes: acquiring an image sequence, the image sequence including a plurality of images of an object in a time dimension; and according to at least one among confidence level change information and depth change information of a keypoint of the object in the plurality of images, determining an occlusion state of the keypoint.