Depth Image Hand Tip Detection Using Inner Product Vector Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In environments with numerous objects, accurately identifying and tracking a desired target in a three-dimensional space is challenging due to factors like changing light sources and high processing loads, leading to delayed response times.
Innovation Solution
An information processor equipped with a depth image acquisition portion and a coordinate point analysis portion that acquires depth images and identifies target positions by analyzing coordinate points within set detection areas, allowing for efficient and accurate position detection in three-dimensional spaces.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If detection process is performed at higher temporal and spatial resolutions for higher accuracy, then measurement precision is improved, but processing load increases and response time deteriorates
Solution Approach 1:
The patent divides the detection space into multiple regions of interest (ROIs) and processes only those regions rather than the entire image. This segmentation allows high-resolution detection to be applied selectively to specific areas containing potential targets, reducing overall processing load while maintaining detection accuracy in critical zones.
Solution Approach 2:
The patent applies different processing resolutions to different regions of the image. High temporal and spatial resolutions are applied only to regions containing detected targets or potential target areas, while other regions are processed at lower resolutions or skipped entirely. This local quality approach maintains measurement precision where needed while reducing processing load elsewhere.
2Measurement precision
If detection process is performed at higher temporal and spatial resolutions for higher accuracy, then measurement precision is improved, but processing load increases
Solution Approach 1:
The patent segments the image into multiple regions and applies high-resolution detection only to regions containing targets or potential targets. This segmentation strategy maintains detection accuracy while significantly reducing the total number of pixels processed, thereby lowering processing load and improving productivity.
Solution Approach 2:
The patent performs detection at high resolution only for portions of the image that contain or may contain targets, rather than processing the entire image at high resolution. This partial action approach applies the necessary processing power selectively, maintaining measurement precision while reducing overall processing load.
3Adaptability or versatility
If numerous objects exist in the shooting space, then adaptability is improved, but difficulty of detecting and measuring increases
Solution Approach 1:
The patent divides the complex scene into multiple regions of interest and processes each region separately. This segmentation reduces the complexity of detecting targets among numerous objects by confining the search to specific areas, making target identification easier while maintaining the ability to handle diverse environments with many objects.
Solution Approach 2:
The patent applies enhanced detection algorithms and higher processing resolution only to regions containing multiple objects or potential targets, while using simpler processing for empty regions. This local quality approach improves target identification difficulty in complex areas without unnecessarily increasing processing load in simple areas, maintaining adaptability to various environmental complexities.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A detection area 112 is set in a three-dimensional space in which a subject exists. When an actual hand enters the detection area 112, coordinate points (white and black dots) represented by pixels making up a silhouette 114 of the hand in a depth image enter the detection area 112. In the detection area 112, a reference vector 126 is set that shows the direction which the hand should face relative to the shoulder as a reference point 122. Then, an inner product between two vectors, a vector from the reference point 122 to each of coordinate points and the reference vector 126, is calculated, followed by comparison between the inner products. Positions of coordinate points whose inner products are ranked high are acquired as the position of tips of the hand.