Gesture Identification Using Depth-Based Region Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In virtual reality technology, existing gesture identification methods face challenges with high computational loads and incorrect hand detection when multiple hands are present in an image, leading to ineffective information input.
Innovation Solution
A method and device that determine the depth of each pixel in images collected by multiple cameras, allowing for the identification of a target region and reducing unnecessary image processing by calculating camera parameters and determining a target depth based on user-specific characteristics or history data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the mobile apparatus analyzes and processes all images collected by the camera to identify gestures, then gesture identification can be achieved, but the computational load becomes very large and processing efficiency deteriorates
Solution Approach 1:
The patent divides the image processing task into two stages: first, depth information is used to segment the image into a target region containing the hand and a non-target region; second, only the target region undergoes detailed gesture analysis. This segmentation approach maintains gesture identification accuracy while dramatically reducing the computational load by processing only relevant portions of the image.
Solution Approach 2:
The patent performs preliminary depth calculation for all pixels in the image before conducting gesture analysis. By pre-processing the depth information and using it to identify the target region in advance, the system prepares the data structure needed for efficient subsequent processing, avoiding the need to analyze entire images and thereby improving processing efficiency.
2Ease of operation
If the mobile apparatus processes images to determine hand position and gesture, then gesture input can be achieved, but when multiple hands are present in the image, the correct target hand cannot be identified
Solution Approach 1:
The patent introduces depth information as an additional dimension to distinguish between multiple hands in the image. By calculating depth values for each pixel and using depth thresholds, the system can identify which hand is within the acceptable depth range for interaction, thereby accurately selecting the target hand among multiple hands and maintaining gesture input functionality.
3Measurement precision
If the mobile apparatus processes the entire image to identify gestures, then all regions are analyzed, but unnecessary computation is performed on regions not containing the hand
Solution Approach 1:
The patent extracts only the relevant target region containing the hand from the entire image using depth information. By isolating and processing only this extracted region for gesture analysis, the system maintains complete gesture detection capability within the target region while eliminating unnecessary computational energy consumption on background and non-relevant areas.
Data Source
Figure 1A~1B
Figure 2~4
Figure 5~7
AI summary
The present invention relates to a method and a device for identifying a gesture. The method includes: determining a depth of each pixel in each of a plurality of images (120, 130) to be processed, in which the plurality of images to be processed are collected by the plurality of cameras (O1, O2), and the depth is configured to at least partially represent a distance between an actual object point corresponding to each pixel and the mobile apparatus; determining a target region in each of the plurality of images to be processed according to the depth; and determining a gesture of a target user according to image information of target regions.