Hand-raising Detection via 3D-to-2D Overhead Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing hand-raising detection systems are ineffective in accurately distinguishing hand-raising actions from other gestures due to limitations in three-dimensional measurement and image processing, particularly in converting 3D data into overhead view images for reliable silhouette extraction and ellipse fitting.
Innovation Solution
A hand-raising detection device that employs a 3D sensor to convert a predetermined space into an overhead view image, using a silhouette extraction unit and ellipse fitting to detect hand-raising actions by analyzing the area ratio between the silhouette image and the fitted ellipse, while also considering the orientation of the face to enhance fingertip detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional three-dimensional measurement and image processing methods are used, then the detection system can operate with simple processing, but the accuracy of distinguishing hand-raising actions from other gestures deteriorates
Solution Approach 1:
The patent transforms 3D spatial measurement data into a 2D overhead view image representation. This dimensional transformation allows the system to view the person's silhouette from above, making hand-raising actions more distinguishable from other gestures by changing the observation perspective. The conversion process creates a top-down projection where hand positions relative to the body become more apparent.
Solution Approach 2:
The patent segments the person's body into distinct regions by fitting an ellipse to the silhouette image. This segmentation separates the head region (identified by the ellipse) from the hand regions, enabling independent analysis of hand positions. By dividing the silhouette into meaningful segments, the system can specifically detect hand-raising actions without being confused by other body movements.
2Reliability
If simple silhouette extraction is used, then the processing is fast and simple, but the reliability of hand-raising detection deteriorates due to false positives
Solution Approach 1:
By converting 3D data to an overhead view 2D image, the system gains a new perspective that reduces ambiguity in gesture recognition. This dimensional change allows the silhouette extraction to be more reliable because hand-raising actions create distinct patterns in the overhead view that are not present in other perspectives, thereby reducing false positives.
Solution Approach 2:
The patent performs preliminary ellipse fitting to identify the head region and establish the body's orientation before detecting hand-raising actions. This preliminary action sets up reference frames and expected patterns that guide the subsequent hand detection process, improving reliability by comparing actual hand positions against predicted positions based on body orientation.
3Measurement precision
If conventional image conversion methods are used, then the conversion process is simple, but the precision of fingertip detection deteriorates
Solution Approach 1:
The overhead view image conversion transforms the detection problem into a 2D plane where fingertip positions can be more accurately determined. In this top-down perspective, the projection of hands and fingers onto the image plane creates clearer separation and positioning information, enabling more precise fingertip detection compared to side or front views.
Solution Approach 2:
The system performs preliminary ellipse fitting to establish the head position and body orientation before detecting fingertips. This preliminary action creates a reference framework that guides the fingertip detection process, allowing the system to search for fingertips in expected locations based on body pose, thereby improving precision.
Data Source
AI summary
A hand-raising detection device includes a converter and a detection unit. The converter performs conversion of a predetermined space including a person into an overhead view image by using a result of three dimensional measurement performed on the predetermined space. The detection unit performs detection of a hand-raising action by using a silhouette image of the person in the overhead view image resulting from the conversion performed by the converter.