Bed Region Detection Using Object Detection and Local Contour Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing bed region detection systems face challenges in accurately detecting the bed area, especially when parts of the bed are obscured by comforters or have linear textures, leading to degraded detection accuracy and increased complexity in large-scale hospital settings where bed installations are frequently moved.
Innovation Solution
A watching assistance system that uses object detection to specify the approximate position of the bed and then extracts the bed contour from a local range, employing machine learning-based discriminators for improved detection accuracy and robustness, simplifying installation and adaptation when the bed or imaging device is moved.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If linear edges are extracted from the image to detect the bed region, then the detection process is simplified, but the detection accuracy is significantly degraded when the bed contour is hidden or when there are many linear edges from bedspreads and surrounding objects
Solution Approach 1:
The detection process is divided into two sequential stages: first detecting the approximate bed position using object detection, then extracting the bed contour from a local range around that position. This segmentation allows the system to benefit from both the simplicity of linear edge extraction and the accuracy of localized contour detection, as the local range extraction is performed only after the rough position is identified.
Solution Approach 2:
Instead of processing the entire image for contour extraction, the system applies contour extraction only to a local range around the detected approximate bed position. This local quality approach reduces the number of linear edges that need to be processed while maintaining detection accuracy in the critical bed region, effectively filtering out irrelevant linear edges from bedspreads and surrounding objects.
2Measurement precision
If manual setting of the bed region is performed, then the detection accuracy is ensured, but the installation burden increases significantly in large scale hospitals with many beds
Solution Approach 1:
The system performs automatic bed region detection without requiring manual configuration by installers. The object detection unit automatically identifies the bed position, and the contour extraction unit automatically defines the bed region boundaries. This self-service capability eliminates the time-consuming manual setting process while maintaining accurate bed region detection, enabling rapid deployment in large scale hospitals.
Solution Approach 2:
The system performs preliminary automatic detection of the bed region during installation, replacing the need for subsequent manual adjustments. By automatically detecting the bed position and extracting the contour in advance, the system ensures accurate bed region definition without requiring installers to spend time on manual configuration and adjustment procedures.
3Measurement precision
If manual setting of the bed region is performed, then the detection accuracy is maintained, but the system becomes complicated to reconfigure when beds are moved
Solution Approach 1:
The system transitions from a static manual configuration approach to a dynamic automatic detection approach. When beds are moved to new positions, the object detection unit automatically detects the new bed position, and the contour extraction unit automatically updates the bed region boundaries. This dynamic capability allows the system to adapt to bed repositioning without requiring manual reconfiguration, maintaining both accuracy and flexibility.
Solution Approach 2:
The automatic detection system serves multiple functions: it initially detects bed regions during installation, continuously monitors for bed position changes, and automatically reconfigures when beds are moved. This universal approach replaces multiple separate processes (initial setup, monitoring, and reconfiguration) with a single integrated automatic detection mechanism, enhancing system adaptability while maintaining accuracy.
Data Source
AI summary
In a watching assistance system that assists watching over a subject on a bed, a region setting unit performs processing of receiving a region setting image captured by an imaging device, specifying an approximate position of the bed from the region setting image by object detection, and specifying a bed region by extracting a contour of the bed from a local range including the approximate position.